Global versus local environmental impacts of grazing and confined beef production systems

In the region assumed for this study (department of Rocha), beef production accounts for 60% of farmers' income and 82% of agricultural land use (MGAP 2002) with an average stocking rate of 0.67 animal units (AU) ha⁻¹.¹ This area was selected because it is representative of Uruguay where beef production accounts for 50% of the farmers, 77% of land use (MGAP 2002) and has an average stocking rate of 0.68 AU ha⁻¹.

The dominant landscape in the area combines hills and slight slopes (3–5%) with valleys with flooding periods during winter. In short distances (1 km) a high diversity of soils can be found, from shallow soils (20–30 cm) in the upper part of hills, to deep (80 cm⁻¹ m) and fertile (6% of soil organic matter) (MGAP 1976). Annual mean precipitation in the region is 1015 mm, with a homogeneous monthly distribution.

Based on previous published literature and expert opinion, we identified two background and three finishing beef production stages (Ferrés 2004, Berretta 2003). The background stage comprises the growth period between 150 and 350 kg; all these systems were based on grazing, either natural rangelands (RL) or seeded pastures (SP) (Berretta 2003, Berretta et al 2000, Risso 1997). Finishing is the stage between 350 and the slaughter weight of 500 kg and is based on RL, SP and feedlot (FL) (Ferrés 2004, Chiara and Ferreira 2012, Berretta 2003, Berretta et al 2000, Risso 1997). The most common systems in the region are five different combinations of these background-finishing stages: RL–RL, RL–SP, SP–SP, RL–FL and SP–FL. We then identified five representative commercial farmers in the region, who kept good records of their management and production through interviews with local extension agents. Through interviews we collected the necessary information to describe the detailed diets, weight gains and land use allocation for one typical cycle of production. The nutritional requirements of each animal were determined using average weight gain, following National Research Council (NRC) (1996) and Agricultural and Forage Research Council (AFRC) (1993) guidelines for British breed steers. Feed intake was assumed as 100% of supplements (grain or hay) offered or forage utilized, assuming 70% of utilization for forage estimated yield for seeded pastures and 60% for rangelands (Holechek et al 1989). Average daily gain was assumed constant for the entire period for each system. Inputs for nutritional calculations were initial and final animal weight, average daily gain, and feed characteristics (type of diet and how it was offered). With this information, nutritional requirements were calculated, as well as the amount of forage and grain required to fulfil those requirements, based on nutritional characteristics of forages and concentrates reported by Mieres et al (2004).

Each system nutritional characteristics and required time to finish the steers are presented in table 1. Crop–pasture rotations in SP systems consisted of two years of crops and three years of pastures. The crops for SP (background) were ryegrass (Lolium multiflorum Lam.) and sorghum (Sorghum bicolor (L.) Moench), and for SP (finishing) the crops were sorghum (Sorghum bicolor (L.) Moench) and rice (Oryza sativa). All seeded pastures were a mixture of fescue (Festuca arundinacea Schreb.) + white clover (Trifolium repens L.) and birds foot trefoil (Lotus corniculatus L.) For feedlot systems continuous cropping of rice and sorghum (Sorghum bicolor (L.) Moench) were considered. All rotations managed soil under no-tillage.

Nutritional requirements were used to calculate the relative area for each system needed to produce the required amount of feed, using national technical coefficients for crop and forage production (table 2). These outputs were the activity data inputs for calculating environmental impacts.

Farmers and farm managers provided the inputs used in one cycle to produce the forages and grains on each system (table 3).

Methane (CH₄), nitrous oxide (N₂O) and carbon dioxide (CO₂) were considered in calculating GHG emissions based on Intergovernmental Panel on Climate Change (IPCC) equations (IPCC 2006). The equations for enteric fermentation, manure management (CH₄), production and distribution of animal's feed (N₂O and CO₂) are presented in table S1 (available at stacks.iop.org/ERL/8/035052/mmedia) and coefficients and emission factors in table 4. Methane emissions from enteric fermentation were estimated based on the energy density (gross energy consumed per day) and the quality of the diet (digestibility and crude protein). Greenhouse gas emissions from manure management are affected by both the animal's nutrition and the effluent treatment system. Nitrous oxide emissions from the production of feed sources include direct and indirect GHG emissions from both the manure and the urine deposited on pasture in grazing systems and from the manure management system in feedlots. Carbon dioxide emissions include fossil fuel combustion machinery used in farming and food distribution.

Extraction, manufacture and transport of inputs, direct and indirect GHG emissions from N fertilization (N₂O), and diesel combustion (CO₂) for feed production were included in calculations. Feedlot systems use diesel combustion for feed distribution.

A tier 2 IPCC (2006) approach and information from Ministerio de Industria, Energía y Minería (MIEM) (2010) were used in calculating diesel and glyphosate emission factors. Methodology described by Spielmann et al (2007) and 9.0 × 10⁻³ kg CO₂-eq Mg km⁻¹ from (Ledgard 2011) was used to calculate emission factors of extraction of raw materials, manufacture, and transport of fertilizers and pesticides. Means of transportation and amounts imported in the last 5 years were obtained from MGAP (2011b). Greenhouse gas emissions derived from seed production were calculated by the simulation of one production cycle and a seed harvesting index from Carámbula (1981).

Carbon stock in soil was assumed to remain constant, as recommended by IPCC (2006). Global warming potential was 1 for CO₂, 25 for CH₄ and 298 for N₂O for 100 years (Forster et al 2007). Greenhouse gas emissions were expressed in kg of CO₂ equivalent per kg of body weight gained by the animals in the background and finishing period (BW) and by carcass weight (CW) considering a slaughter yield of 58.3% (steer's average for year 2011) (INAC 2012).

Fossil fuel energy consumption was estimated using the coefficients in table 5 and farm-specific data from tables 2, 3 and 12 (yields, inputs and total area per animal). The energy use per kg of live weight gained was estimated over one growing cycle and expressed in MJ kg body weight⁻¹ (BW).

Soil erosion rates for each system were estimated using EROSION 5.0 (García Préchac et al 2005), a model based on the Universal Soil Loss Equation (Renard et al 1994) and its later Revision (USLE/RUSLE) adapted and calibrated for Uruguay soils.

The USLE is a multiplication of factors, derived from climate (erosivity), topography (erodability, length and inclination of slope) and management (use and management and mechanical activity).

Rain erosivity factor is the rainfall energy applied to soil, derived from local meteorological data (J ha⁻¹). Soil erodability is the average soil lost by unit of erosivity on bare soil (Mg J⁻¹). Length slope factor is the relation between erosion at a given length and the one occurring at 22.1 m, when all other factors remain constant. The slope inclination factor is the relation between erosion at a given slope and the one occurring at 9% slope, when all other factors remain constant. The use and management factor is the relation between soil erosion under a set system and the erosion that would occur from a soil under bare soil, when all other factors remain constant. The mechanical activity factor is the relation between soil erosion under a determined mechanical activity and that occurring under a standard tillage in favor of the slope, when all other factors remain constant.

The same location (department of Rocha, SE of Uruguay) was considered for all systems, as well as soil type (Brunosol subeutrico típico—Typic Argiudolls) (MGAP 1976) and topography with slopes of 3% and 100 m long. These erosion rates (in kg ha⁻¹) were multiplied by the area needed to produce feed to finish one animal to calculate total erosion (in kg), and expressed as kg of soil loss per kg of animal BW.

Nitrogen (N) and phosphorus (P) nutrient imbalance ratio (NIR) was calculated for each system using the methodology proposed by Koelsch and Lesoing (1999). This indicator is a ratio between nutrient outputs and inputs used in the system. Values over one represent nutrient surpluses (higher surpluses represent higher risks of water contamination) and values under one represent that nutrients are being exported at a higher rate than incorporated in the system.

Nitrogen inputs considered fertilizers and nitrogen biological fixation from legume pastures while the body weight gained in the period was considered as output. On rangelands nitrogen biological fixation is not accounted since the presence of legumes in the botanical composition of this communities are relatively low (Nabinger et al 2000). Phosphorus inputs considered fertilizers as inputs and the body weight gained in the period as output.

Coefficients for these calculations are presented in table 6.

Table 6.  Coefficients and equations used in nutrient (N and P) imbalance calculations. (Note: NBF: nitrogen biological fixation; EBW: empty body weight (55% of the animal's live weight, assuming British breed steers).)

Coefficient	Value/equation	Reference
Animal's N content	$N=\frac{(0.2 3 5\times \mathrm{EBW}-0.0 0 0 \hspace{0.167em} 1 3\times \mathrm{EBW}2-2.4 1 8)}{6.2 5}$	NRC (1996)
% N NBF of Legume pasture (25–90%) 1∘ year of production	18% of harvested weight	Koelsch and Lesoing (1999)
% N NBF of Legume pasture (25–90%) 2∘ year of production	36% of harvested weight
% N NBF of Soybeans	40% of harvested weight
Animals P content	0.69% of EBW	NRC (1996)

We based our calculations on an index proposed by Viglizzo et al (2006), which considers each pesticide toxicity values (LD-50), solubility (Ksp), adsorption (Koc) and the time required for the pesticide to decline to 50% under field conditions. These values were obtained from the pesticide properties database (PPDB) (2009). Products used on each system and its coefficients are presented in table 7.

All coefficients but LD-50 are translated to a scale with values from 1 to 5 proposed by Webber (1994) (table S2 available at stacks.iop.org/ERL/8/035052/mmedia). These values are multiplied by the doses applied by farmers and the relative area of the plot in the farm (equation (1)).

Pesticide contamination risk calculation (Viglizzo et al 2006)

$$\begin{eqnarray} \mathrm{Risk}&=&\frac{1 0 0 0}{\mathrm{LD}\hspace{0.167em} 5 0}\times \left [\frac{\mathrm{Ksp}}{2}+\mathrm{Koc}+{T}^{1/2}\right ]\nonumber\\ &&\times ~\mathrm{dose}\times \text{relative area} \end{eqnarray} \tag{ 1 }$$

where: LD-50: lethal dose-50 (mg kg⁻¹); Ksp: solubility (g g⁻¹); Koc: adsorption (g g⁻¹); T^1/2: time for 50% decomposition under field conditions (days).

Aquifer recharge coefficient was not considered since we assumed that the type of soil, distance to aquifer recharge points and water bodies is the same for all fields of all farms.

Even though this is a relative index it is useful to compare between farms, mainly when the final product is the same but relying on different management and technologies to accomplish that goal.

A threshold is the limit of a certain variable that, if breached, it is impossible or very difficult to recover from, because the system changes its basic function (Walker et al 2004). For the analyzed variables, only soil erosion has a definition that could be translated into a threshold. That is the 'tolerance' value (T), defined as the maximum soil erosion rate that can happen matching or being less than the rate of soil formation (Barrow 1991). In the case of Uruguay, this rate has been defined in 7 Mg ha yr⁻¹ (Puentes 1981). For GHG and energy consumption we compared our results with previous life cycle assessment (LCA) analyses. Nutrient balances and pesticide risk of contamination were compared with the results obtained by the authors who proposed each methodology (Koelsch and Lesoing 1999, Viglizzo et al 2006), since there are not many reported values beyond those analyses.

Greenhouse gas emissions are lower as systems intensify, with the highest value in RL–RL and the lowest in SP–FL (table 8). Systems with feedlots had lower values, even though one is combined with rangelands. For all systems CH₄ was the most important GHG, followed by N₂O and CO₂. This is explained by the source of GHG emissions, where enteric fermentation is higher than 46% of GHG emissions in all cases.

Nitrogen fertilizer, crop residues and inputs represent more than 10% of GHG emissions only in the SP–FL system, the system with more intensive use of inputs.

Our results are within the highest and lowest values previously reported (table 9). Our highest value (RL–RL) is the second highest one after the Ogino et al (2004).

We performed a sensitivity analysis on the N₂O emission factors (EF₃,EF₄ and EF₅) since these are the ones with the highest range (IPCC 2006). To perform these calculations we used the lowest and highest values of the range for each emission factor and analyzed the variation with the average or given value default value given by IPCC (2006). Results are shown in table 10. The highest impact for all systems was changing the EF₃ either to the lowest or highest value. On average the lowest EF₃ decreased 15% the GHG emissions and the highest 44%.

There does not seem to be a pattern following the different production systems or a difference between pure grazing systems or those combined with feedlots. Although there is a difference in magnitude of the change between systems, these differences are not big enough to change the order of the GHG emissions of the different systems when considering the same scenario.

Energy consumption is greater as systems rely more on inputs to produce 1 kg or BW. Background-finishing under natural rangeland systems (RL–RL) did not use any fossil fuel energy, relying 100% on solar energy. Seeded pastures systems burned fossil fuels for pastures and crop production both in field and for fertilizers and pesticides industrial production. The most intensive system (SP–FL) required 2.5 times more fossil energy than the lowest input-use system and 1.5 times the second highest input-use system to produce one kg of BW (table 11).

Our results are different than previous reports from Capper (2012), Pelletier et al (2010) and Peters et al (2010) who reported less fossil fuel energy consumption from feedlots than grazing systems. Differences may arise firstly because our study considered rangeland systems, where no fossil fuel energy is used. Capper (2012) mentions that the greater use of fossil fuels derives from cropping and harvesting of forages. In the case of seeded pasture systems the use of fossil fuel energy for cropping is mostly in the first year of seeding the pasture because of the use of perennial species that produce enough forage for three or more years. Furthermore, animals graze pastures directly, without the intervention of machinery, which also reduces the consumption of energy. Therefore, in our conditions and pastures, grazing systems have less energy consumption than feedlots. Comparison of our results with previous published data is presented in table 12.

Erosion rate increases as cropping activities become a higher proportion of the area required for producing 1 kg BW. Systems where livestock production relies on supplements produce more soil erosion. Erosion rates were very different between systems when expressed by unit of area, but differences become smaller when these numbers are converted in Mg kg BW⁻¹, with Rangeland systems with the lowest values and increasing to systems with higher use of supplements (table 13). Although this different impact on soil erosion, none of the systems crosses the threshold defined for this variable (7 Mg ha⁻¹) (Puentes 1981).

Nutrient imbalance ratio does not consider the amount of nutrients provided by the soil and the rain. Outputs such as leaching, runoff, and erosion are not measured. It can be inferred that higher surpluses could lead to higher risks of leaching and runoff, e.g. water contamination.

Rangeland systems had the smallest amount of inputs for N and P, while SP–SP had the highest inputs for N, where the most important source is nitrogen biological fixation by legume pastures (table 14). Fertilizers were the main source of N and P for feedlots, with the highest levels of P inputs.

Since output of nutrients is the same among systems, the NIR increases as systems use more inputs (table 14). For nitrogen, RL–RL exports more N that the total inputs and SP–FL had the highest ratio, with 4.8 kg of N going in the system for each kg of N going out as BW. These results are lower than those reported by Whitehead (2000) who reported values of 7.7 and 14.3 for extensive and fertilized grasslands, respectively.

Phosphorus NIR follows the same path, with SP–FL with the highest surpluses of P and RL–RL with a negative balance, since 70% of the exported P as live weight is not incorporated as input. The other systems NIR where over 1.6 and 1.5 reported by Koelsch and Lesoing (1999); while SP–SP and SP–FL where over the 3.1 value reported for feedlots by Koelsch (2005).

From these results two opposite situations arise: firstly, systems with nutrient surpluses (from 2.4 to 4.8 times more inputs than outputs) which could lead to leaching and runoff, hence higher risk of water contamination. On the other hand, rangeland systems could be undermining P and N soil reserves since inputs don't compensate for the extraction.

Since rangeland systems do not use pesticides at all, the relative risk of contamination is zero. All other systems used pesticides and results were 2.4, 6.3, 12.2 and 21.2 for RL–FL, RL–SP, SP–SP and SP–FL. The risk increases as we move to systems with higher use of inputs (figure 1). Systems using pesticides are clearly differentiated mainly by doses and the relative impact of pesticides used to produce fodder, since different parameters (Ksp, Koc, T^1/2 and LD50) compensate each other when comparing between systems (table 7, and S2 available at stacks.iop.org/ERL/8/035052/mmedia). Our results are within the limits of those reported by Viglizzo et al (2006), who calculated this index for 120 farms in the Argentinean Pampas, finding values between 0 and 44.

Zoom In Zoom Out Reset image size

Analyzing sustainability of systems should incorporate a broad concept of variables to avoid over-simplification. On the other hand, using too many indicators could generate confusion when discussing results.

We described two global environmental problems (climate change and fossil fuel derived energy consumption) and four variables driving local environmental problems (soil erosion, N and P balance and risk of pesticide contamination). While this is a broad classification, it could help to guide stakeholder's decisions to promote or regulate environmental impact of livestock activities.

There is a trade-off among the six environmental variables analyzed. As systems intensify the use of inputs and increase productivity per unit of resource (land, animals, capital), they perform better in terms of GHG emissions per kg of LW (Capper 2012, Ogino et al 2004) but worse in the other indicators (figure 1).

This trade-off between global and local environmental variables calls for further research in order to define priorities for development. While in order to mitigate a global issue like climate change it could be reasonable to promote beef production in feedlots, in order to mitigate local environmental problems like water contamination through nutrient leaching and pesticides, as well as soil erosion grazing systems are preferred. Also, although rangeland systems seem to be more environmentally friendly systems, they could be undermining soil nutrients in the long term if there is no reposition of N and P extracted.

Quantitative thresholds can give a guide to design more sustainable production systems but the development of these values is still incipient. This analysis gives hints for stakeholders in order to define policies related to food and environment, although further dimensions need to be considered (e.g. biodiversity, water footprint).