Legacy effects of individual crops affect N2O emissions accounting within crop rotations

Uruguay is pursuing renewable energy production pathways using feedstocks from its agricultural sector to supply transportation fuels, among them ethanol produced from commercial technologies that use sweet and grain sorghum. However, the environmental performance of the fuel is not known. We investigate the life cycle environmental and cost performance of these two major agricultural crops used to produce ethanol that have begun commercial production and are poised to grow to meet national energy targets for replacing gasoline. Using both attributional and consequential life cycle assessment (LCA) frameworks for system boundaries to quantify the carbon intensity, and engineering cost analysis to estimate the unit production cost of ethanol from grain and sweet sorghum, we determined abatement costs. We found 1) an accounting error in estimating N2O emissions for a specific crop in multiple crop rotations when using Intergovernmental Panel on Climate Change(IPCC) Tier 1 methods within an attributional LCA framework, due to N legacy effects; 2) choice of baseline and crop identity in multiple crop rotations evaluated within the consequential LCA framework both affect the global warming intensity (GWI) of ethanol; and 3) although abatement costs for ethanol from grain sorghum are positive and from sweet sorghum they are negative, both grain and sweet sorghum pathways have a high potential for reducing transport fuel GWI by more than 50% relative to gasoline, and are within the ranges targeted by the US renewable transportation fuel policies.


Introduction
There has been much interest over the last decade in developing and scaling renewable energy to address both climate change and energy security in countries around the world (Kammen, 2006). Rationale for developing renewable energy varies by country and economy, but a strong case for building domestic supply capacity is to support rural economic development. Policies for pursuing renewable energy for decarbonizing transportation favor the development of domestic commercial biofuels as these also help to invest in local economies and reduce trade deficits related to foreign oil dependence. In North America, these policies have been designed to follow life cycle assessment (LCA)based greenhouse gas (GHG) accounting (U.S. Congress 2007;US EPA 2010). Countries around the world are developing similar policies in an effort to spur their local economies, and in some cases, also to mitigate climate change. In Uruguay, domestic biofuels are in production and are poised to expand production of domestic energy resources. Uruguay's governmentowned oil company, ANCAP, has invested in domestic production of ethanol from sorghum crops, but to this point, they have not evaluated the climate impacts of their biofuel program. Our goal was to develop a life cycle framework to evaluate their current biofuels from sorghum, using this system approach to inform policy to understand and guide the life cycle performance of Uruguay's domestic biofuel investment using metrics such as GHG accounting.
Energy is an important driver for economic growth, and for many developing nations, its consumption is projected to increase to improve standards of living. In Uruguay, demand for energy has increased 70% over the last 10 years, with the industrial sector being the highest consumer (~76 PJ), followed by the transportation (49 PJ) and residential and commercial (46 PJ) sectors (DNE, 2013). Near-term options for meeting energy demand have been met by increased import of petroleum and a small, but expected to increase in import of natural gas since Uruguay lacks domestic reserves of fossil energy resources. There are no reserves of coal, petroleum, or natural gas, and thus, historically any use of such resources for thermal, electrical, and transportation has relied on import from the global market. In 2013, fossil energy accounted for 41% of energy consumed, 40% of this was comprised of petroleum, and 1% natural gas, while renewable energy accounted for 59% of energy consumed, with 39% coming from waste, 33% hydroelectric power, 19% from wood, 6% from wind, and 3% from biofuels. The transportation sector, almost exclusively dependent on liquid fuel supply, largely relies on diesel and gasoline refined from imported crude oil, with some domestic production of fuel ethanol and biodiesel from domestic agricultural products accounting for~1.5% of total energy (DNE (Direcci on Nacional de Energ ıa), 2013). Biomass from forest and agricultural sectors is an important source of renewable energy to support effort to address climate change. Near-term energy policy in Uruguay is focusing on developing liquid fuel markets from agricultural feedstocks and electricity from forest biomass, which have been shown to reduce GHGs associated with energy production and use .
LCA approaches are being pursued to best guide decisions in both energy sectors; GHG accounting and cost abatement (e.g., Pourhashem et al., 2013) are valuable metrics to help understand the effectiveness of policy investments, in particular when government is investing in carbon abatement strategies. Transportation fuel supply must be compatible with the vehicle fleet and new vehicles entering the market. Thus, ethanolblended fuels can be scaled up to the blend wall for the existing fleet of gasoline-fueled vehicles and up to 85% for flexible fuel vehicles.
Policy around low-carbon transportation fuels in the United States (CARB 2010; US EPA 2010) uses a combination of attributional LCA (ALCA) (ISO 2006) and consequential LCA (CLCA) (Ekvall & Weidema, 2004) methods, the latter being used primarily to quantify carbon emissions from indirect land-use change induced by changes in commodity markets (Searchinger et al., 2008;Plevin et al., 2015). CLCA is also used to guide system boundary rules that test a new product or production strategy against a conventional (baseline) product or production strategy. For example, Sarkara & Miller (2014) used CLCA boundary rules for assessing changes in water quality when introducing switchgrass into agricultural landscapes for bioenergy production. Similarly, both Pourhashem et al. (2013) and Adler et al. (2015) used CLCA boundary settings to test the effects of soil carbon management within biofuel systems and use these settings to estimate the soil GHG emissions of nitrous oxide and soil carbon change. Plevin et al. (2014) argue that ALCA fails to account for the critical changes within the economy that may result from implementing climate change policy, especially for biofuels. These changes can lead to error when comparing the global warming intensity (GWI) of a biofuel relative to the fossil fuels they aim to displace. The authors conclude that although CLCA can account for change, it also introduces results that are scenario dependent and uncertain when used to evaluate biofuel policy.
In addition to scenarios which define how cropping systems may change with the introduction of a bioenergy crop, there is also the interaction of crops within a rotation which can affect the GHG emissions for a given crop within both the ALCA and CLCA frameworks. In this analysis of bioenergy production in Uruguay, there is not only the increase in planting of grain and sweet sorghum, but the change in crop identity in the rotation due to external economic and regional factors.
Soil biogeochemical changes across crop rotations leading to GHG emissions of N 2 O and CO 2 from soils are not always precisely captured when using CLCA or ALCA due to limitations in the estimation methods used. The Intergovernmental Panel on Climate Change (IPCC) has developed guidelines to estimate N 2 O emissions from cropped and grazed soils, as well as soil organic carbon (SOC) stock changes (de Klein et al., 2006). Explicit in the guidelines are tiers reflecting methodological complexity with Tier 1 methods based on default emission factors being the simplest, Tier 3 methods employing the complex process-based models, and Tier 2 methods being intermediate. Process-based models (Tier 3) have matched measured N 2 O emissions more closely (Del Grosso et al., 2008) and represented soil carbon changes better (Del Grosso et al. 2016) than IPCC Tier 1 methods. Furthermore, as hypothesized in this study, when considering complex rotations, Tier 1 methods do not capture legacy effects of nitrogen management from prior crops in a rotation because they assume that all applied N is cycled within a year. This could lead to underestimates of N 2 O emissions with Tier 1 methods due to carryover of N from one crop in a rotation to another, a source of N not accounted for in Tier 1 in contrast with Tier 3 N 2 O estimation methods.
Given the challenge of appropriately capturing changes in agricultural GHG emissions when a biofuel policy and the crops that meet the policy are introduced, and the potentially divergent estimation outcomes from Tier 1 and Tier 3 soil GHG accounting, our objective in this article was to evaluate the GWI of newly introduced biofuel pathways in Uruguay's varied agricultural crop rotations using Tier 1 and 3 approaches and CLCA and ALCA frameworks. We explore major differences in direct GHG emissions for biofuels introduced into complex agricultural rotations and posit that decision outcomes from life cycle inventories defined by CLCA and ALCA frameworks can lead to very significant soil N 2 O accounting omissions, which will impact how the biofuel program ranks and rates petroleum alternatives. While CLCA encompasses a wide spectrum of changes introduced by a new product or policy, including especially economic effects resulting from market changes (Anex & Lifset, 2014;Hertwich, 2014;Plevin et al., 2014;Suh & Yang, 2014), our goal was to investigate only biophysical changes in soil N 2 O emissions resulting from the introduction of a crop within a multicrop system using CLCA because it is not possible to capture such effects with ALCA methods used in prior literature (e.g., Spatari and MacLean 2010). We combine GHG accounting with cost analysis to investigate the abatement cost of future bioenergy supply in Uruguay and evaluate the environmental and cost effectiveness of their major investments into grain sorghum and sweet sorghum biorefinery capacity. The analysis aims to explore the range of differences in biofuel life cycle GWI when using consequential and attributional system boundaries, IPCC Tier 1 and Tier 3 approaches to estimate soil GHG emissions, and agricultural rotations scenarios that include and exclude pasture land, which constitutes a major fraction of agricultural land in Uruguay.

Site selection and description
Our analysis focused on two locations in Uruguay where ethanol is being produced from agricultural feedstock, Paysand u in western and Bella Uni on in northern Uruguay (Fig. 1). The site near Paysand u produces~70 000 m 3 ethanol annually from grain sorghum, requiring 156 600 Mg of grain sorghum annually (Table 1). With the regional average grain sorghum yield of~4 Mg ha À1 , the biorefinery requires~40 000 ha of land annually. The site near Bella Uni on produces ethanol from sugarcane and sweet sorghum. They have an annual production capacity of~30 000 m 3 ethanol annually,~95% from sugarcane, and~5% from sweet sorghum. Although these locations have similar climate without a large gradient across the region (Table 2), different crops are produced in these regions due to differences in infrastructure, such as distance to ports, and sugarcane mills in the north. The cropland in Paysand u is more dominated by grain crops, whereas in Bella Uni on, commodity crops are less common, and forage crops more common.

Life cycle assessment
The life cycle assessment (LCA) was conducted in two phases as prescribed by ISO 14040 (2006) procedures. First, all sources of greenhouse gases were tabulated in a life cycle inventory (LCI) analysis and then the contribution of the sources of greenhouse gases on climate was determined by converting the inventory to CO 2 equivalents (life cycle impact assessment, LCIA).
The DayCent model was used to quantify changes in soil organic carbon (SOC), N 2 O emissions, and NO 3 leaching over the crop production cycle. The DayCent biogeochemical emissions were incorporated into the life cycle inventory (LCI) model (Adler et al., 2007. DayCent model description. The DayCent biogeochemical model (Parton et al., 1998;Del Grosso et al., 2001), a daily timestep version of the CENTURY model (Parton et al., 1994), was used to estimate crop yields and evaluate changes in soil organic carbon (SOC) and soil N 2 O emissions for the LCA. Using daily weather, soil-texture class, and land-use inputs, DayCent simulates crop production, soil organic-matter (SOM) transformations, soil water and temperature dynamics, tracegas fluxes, and other ecosystem processes. Plant growth is controlled by nutrient and water availability, temperature, and cultivar-specific characteristics such as phenology, N concentration of biomass components, and maximum growth rate. SOM dynamics are a function of the quantity and quality of biomass inputs, water, temperature and nutrient limitation, tillage intensity, and soil properties related to texture. The model simulates soil N 2 O emissions from nitrification and denitrification, as well as CH 4 oxidation in drained soils. The ability of DayCent to simulate NPP, SOC stock changes, N 2 O emissions, and NO 3 leaching has been tested with data from various native and managed systems (Del US EPA 2013). DayCent has been shown to reliably represent plant growth and GHG fluxes for different biofuel cropping systems, and the model has been successfully applied at the site (Adler et al., 2007) and regional levels .
Daily weather data for Paysand u and Bella Uni on Uruguay required to drive DayCent were acquired from the nearby INIA weather station in Salto Grande (31°S 16 0 22″, 57°W 53 0 27″), 120 km from both biorefinery locations. Soil texture data representative of the locations were obtained from Altamirano et al. (1976). The representative soil in Bella Uni on for sweet sorghum was a Typic Hapludert clay soil of the series Itapebi Tres Arboles, while in Paysand u for grain sorghum, it was a Typic Argiudoll clay loam of the series San Manuel. Both soils are common throughout the cropped areas of the Pampean region of Uruguay and Argentina, very productive, high fertility and poorly drained especially the Hapludert, although poor drainage is usually not a limiting factor because these soils are located in rolling landscapes. Soil physical and hydraulic properties needed for model inputs were calculated from texture class and Saxton et al.'s (1986) hydraulic properties calculator (available online at http://www.bsyse.wsu.edu/sax ton/soilwater).
Model outputs are sensitive to current SOC levels, which in turn are influenced by previous vegetation cover and land management. To acquire reasonable modern SOC levels, about 1950 years of native vegetation followed by plowing and about 60 years of cropping were simulated. Plow out of native grazed grasslands was assumed to occur in the year 1951. Historically accurate cropping systems were simulated, and improved cultivars, fertilizer applications, and tillage intensity were introduced at appropriate times. From 1952 to 1975, corn-wheat rotations were common, no N fertilizer was applied, and conventional tillage was used. From 1976 to 2007, soybean was introduced and included in the corn-wheat rotations and N fertilizer applied; in response to soil degradation, pasture was included in the crop rotation and no-tillage was adopted (Gar . The pastures contain a mixture of grasses (fescue, ryegrass) and legumes (white clover, red clover, sometimes alfalfa) with a low C/N ratio. The modeled rotations were developed by regional experts and are described in Table 3. They include a  ‡Total amount of feedstock required as stalks for sweet sorghum and sugarcane and grain for grain sorghum.
baseline rotation which describes a typical crop rotation prior to expansion of bioenergy crop production in the area, and a paired bioenergy crop rotation. Both the baseline and bioenergy crop rotations included scenarios with and without a three-year pasture component, for both grain and sweet sorghum. The grain sorghum scenario without pasture and the sweet sorghum scenario with pasture would probably be more typical for the regions, but a change in commodity prices could move rotations toward or away from increased grain or pasture in the rotation. DayCent simulations were run for 36 years following initiation of these rotations. Outputs for carbon (C) and nitrogen (N) fluxes between the atmosphere, vegetation, and soil were then used as inputs into the crop production LCI model to estimate the global warming intensity (GWI) of feedstocks for energy use . The direct N 2 O was the mean annual N 2 O emissions over the simulation period. To calculate indirect N 2 O, we combined DayCent outputs for NO 3 leached and N volatilized with IPCC (de Klein et al., 2006) methodology. IPCC (de Klein et al., 2006) methodology assumes that 0.75% of NO 3 -N leached is eventually denitrified to N 2 O-N in water ways and that 1% of volatilized N (NO x + NH 3 ) is deposited on soil and converted to N 2 O. N 2 O emissions were converted to CO 2 e by assuming that its global warming potential is 298 times that of CO 2 on a mass basis (Forster et al., 2007).
IPCC Tier 1 greenhouse gas estimations. The Tier 1 method for soil N 2 O emissions assumes that fixed portions of N additions to soil from fertilizer, manure, and crop residues not removed during harvest operations are converted to direct and indirect emissions. We used the default direct N 2 O factor of 1% for synthetic fertilizer and crop residues for the cropped portion of the rotations and the default 2% for manure deposited by grazing animals during the pasture portion of the rotations. We also used the default indirect factors from IPCC (de Klein et al., 2006) and summed to obtain total N 2 O emissions. For SOC stock changes, the Tier 1 method estimated an initial SOC stock of 8800 g C m À2 given the soil types and climate for the sites in Uruguay. We then applied the default SOC stock change factors based on land-use change as described in IPCC (de Klein et al., 2006) and calculated annual SOC changes by assuming that SOC stocks would obtain equilibrium after 20 years of consistent land use.
Description of LCI and LCIA. DayCent output data on changes in SOC and N 2 O were combined with a life cycle inventory (LCI) model for the selected biomass feedstocks converted to ethanol. Crop yield output from DayCent was expressed as g C m À2 yr À1 and converted to MJ ha À1 assuming biomass feedstock was 43.5% carbon (Brown, 2003), ethanol yield was 442 and 216 L ethanol Mg À1 feedstock for grain and sweet sorghum, respectively (Table 4). The lower heating value of ethanol (21.2 MJ L À1 ) was used for all life cycle calculations. The LCI model follows ISO 14040 (2006) procedures and includes the fuel production (feedstock production, transport, fuel conversion, fuel distribution) and use (combustion) cycles (Fig. 2). Life cycle GHG emissions include transportation and field application of nutrients and farming operations specific to the crop rotation. Coproducts compared among the ethanol pathways were treated using system expansion crediting (ISO, 2006) and those credits depended on the rotation. For example, in the case of grain sorghum, the distillers dried grains and solubles (DDGS) coproduct was assumed to displace soybean meal on the market similar to other DDGS coproducts derived from starch-based dry grind processes. In the case of sweet sorghum, the annual production of surplus electricity (1452 MWh)  ‡The first three years of the rotation are followed by three years of pasture.
produced was assumed sold to Uruguay's electricity market and would displace marginal fuel sources for electricity within the existing electricity mixture, which consist of fuel oil that is used on top of hydroelectric and biomass sources of baseload power. Life cycle impact assessment (LCIA) was undertaken following ISO 14040 (2006) standards to estimate the GWI of the ethanol produced and compared to the GWI of gasoline (93 g CO 2 e MJ À1 ) as documented by US EPA (2010) and includes end use in a light-duty internal combustion vehicle. The GWI of feedstock production was calculated using 2007 IPCC 100a weighting factors for the individual greenhouse gases (Forster et al., 2007) and summed over the cradle-tofarm-gate life cycle. The LCI model for ethanol produced from grain and sweet sorghum was constructed using SimaPro 8 software (PRe' 2015) using existing datasets for select agricultural operations (e.g., feedstock harvest, nutrient replacement) that were parameterized with data collected in Uruguay, and biorefinery operations data from Uruguayan industry (Alur and ANCAP), supplemented with data from literature (e.g., Gnansounou et al., 2005;Nghiem et al. 2011) that were parameterized for biorefinery scales in Uruguay. Life cycle GWI contributions from crop production, biorefinery operations, transportation, and vehicle in-use emissions for 100% ethanol produced from sweet sorghum (Table S1) and grain sorghum (Table S2) are in the Supporting Information.

Description of cost and GHG abatement analysis
To test the performance of abating GHGs through introducing grain and sweet sorghum into Uruguay's transport fuel market, the majority of which is owned by the Uruguayan government, we calculated the GHG abatement costs for displacing gasoline.
Production costs. We estimated the unit production cost of ethanol from grain and sweet sorghum. For grain sorghum to ethanol, costs included the capital investment of $180 million, depreciated over 10 years using straight line depreciation; feedstock and operating variable costs; and fixed variable costs (labor, general overhead, maintenance, and insurance and taxes) using data for corn dry grind facilities (Iowa State University 2015). Grain sorghum feedstock costs were taken from 2015 commodity prices (USDA-NASS 2015). The cost of producing sweet sorghum to ethanol did not include depreciated capital equipment because existing facilities from the Bella Union sugarcane processing plant are used. The feedstock production cost in Uruguay was calculated to be $1143 ha À1 , which is slightly higher than costs reported for U.S. production in Texas, $712 ha À1 (Amosson et al., 2011), and North Carolina, $1023 ha À1 (Veal et al., 2014), and the cost of transporting the feedstock to the biorefinery was estimated to be $7.70 Mg À1 , assuming a transportation distance of 70 km and freight transportation cost of $0.11 Mg-km À1 (Inter-American Development Bank 2015). Operating variable costs and fixed variable costs (labor, general overhead, maintenance, and insurance and taxes) were estimated using data from Gnansounou et al. (2005) and scaled to the Bella Union capacity of 30 000 Mg yr À1 sorghum feed. All costs were assumed to be in 2015 US$. A sensitivity analysis on costs was performed to estimate the range of ethanol unit production costs and used for the abatement cost calculation. Tables S3 and S4 in the Supporting Information summarize the production cost and sensitivity analysis parameter settings for ethanol production from grain sorghum and sweet sorghum, respectively. Abatement costs. We estimated the GHG abatement cost of grain and sweet sorghum ethanol pathways relative to gasoline using Eqn 1 and included low and high ranges to evaluate cost sensitivity.
Data on the assumed average, low, and high costs for grain sorghum to ethanol and sweet sorghum to ethanol are summarized in the Supporting Information Tables S1 and S2. The average wholesale price of gasoline by refiners from January 2013 to June 2015 from (EIA 2015) was used ($0.56 L À1 ; $20.64 GJ À1 ), and the GWI for gasoline was assumed to be 93 g CO 2 e MJ À1 (US EPA 2010).

Crop yields and available crop residue
Grain and sweet sorghum yields were similar in both rotations with and without pasture (Table 4). Although the ethanol yields were more than double per unit yield for grain sorghum, dry wt. yields (stem vs. grain) were more than 2.5 times greater for sweet sorghum; the resulting ethanol yields per hectare were >20% higher for sweet sorghum.

Greenhouse gas emissions and nitrogen dynamics
Greenhouse gas emissions varied with both crop and rotation (Table 5). Total N 2 O emissions were higher than the baseline for both sweet and grain sorghum; a small reduction in nitrate leaching led to lower indirect N 2 O emissions, but these were small relative to direct N 2 O emissions. The change in both N 2 O emissions and soil carbon were lower for both the sweet and grain sorghum rotation with pasture relative to the rotation without pasture.
Cumulative N 2 O emissions varied with crop and rotation. There was a legacy effect following the pasture, whether that period was fallow (Fig. 3) or had crops GHG abatement cost ¼ À Unit production cost of ethanol À Unit production cost gasoline Unit GWI ethanol À Unit GWI gasoline ð1Þ planted (Fig. 4), N 2 O emissions were higher. The effect of pasture on the intensity of N 2 O emissions following pasture increased with proximity; N 2 O emission intensity was higher for oats and wheat which were nearer in time to pasture than sweet sorghum or soybeans, but were still higher for these crops in rotations with pasture. Nitrogen inputs and outputs varied between grain and sweet sorghum and the presence of pasture in the rotation (Table 6). More N fertilizer was added to the grain sorghum rotation, whereas N fixation varied with the presence of pasture and frequency of soybean in the rotation. However, both N mineralization and manure/urine inputs were highest in rotations with pasture. Nitrogen output was highest in rotations with grain sorghum, while grazed N and gaseous N losses where highest in rotations with pasture. Pastures were an important source of internally cycled N leading to large difference in net N input.

Greenhouse gas emissions on a life cycle basis
Both the direction of change between the rotations and quantity varied within and between ALCA and CLCA methods for Tier 1 and Tier 3 methods. Using ALCA methods, Tier 3 N 2 O emission estimates were higher

Sweet sorghum
Grain sorghum Fig. 2 System boundary for Uruguay's sorghum-to-ethanol pathways. Abbreviations: transportation (T); distribution (D); chemical inputs included GHG emissions associated with production of both fertilizers and pesticides; Ag machinery included GHG emissions associated with fuel use from tillage, planting, chemical application, and harvest; N 2 O emissions were from soil and were always positive in contrast to SOC (soil organic carbon) which increased or decreased depending on the rotation. *DayCent simulated crop yields, both grain (grain sorghum) and stem (sweet sorghum) yields are expressed as dry wt. Stem yields were about 75% of total biomass; grain, leaves, and 16.7% of stems for sweet sorghum were returned to the soil at harvest as occurs in commercial operations. †Sweet and grain sorghum biomass included stems, leaves, and grain. ‡Amount of biomass returned to the soil with harvest. §Ethanol yields are expressed on a dry wt. basis.
than Tier 1,~1.7 to >2.3 times higher for sweet sorghum and~1.5 to 1.65 times higher for grain sorghum (Table 7). While N 2 O emissions were similar between rotations within both Tier 1 and Tier 3 methods for grain sorghum, for sweet sorghum, they were lower with pasture for both Tier 1 and Tier 3 methods. ALCA methods only consider N 2 O emissions during the growth of the target crop (i.e., grain and sweet sorghum), in contrast to CLCA which considers change.
With CLCA methods considering the change from a baseline, for grain sorghum N 2 O emissions increased with pasture in the rotation, while for sweet sorghum N 2 O emissions decreased, and Tier 3 were higher than Tier 1 estimates (Table 7). All crop rotation scenarios produced ethanol with a lower GWI than gasoline (Table 8). Using ALCA methods, GWI estimates for Tier 3 were higher than Tier 1, while using CLCA methods, they tended to be lower. All but three crop rotation scenarios reduced the GWI by more than 50% compared with gasoline, and the three exceptions still reduced the GWI by more thañ 40%. Sweet sorghum mostly had a lower GWI than grain sorghum using CLCA but higher using ALCA methods.

GHG abatement costs
On a life cycle basis, all sugar/starch crops for ethanol production maintain a low life cycle GWI largely due to the already low GWI of Uruguay's electricity grid, which is comprised of on average 50% renewable/lowcarbon energy (Fig. 5). In particular, the sweet sorghum pathways, whether assessed using consequential or attributional frameworks had consistently lower life cycle GWI compared to grain sorghum pathways. Moreover, the cost estimates (Tables S3 and S4) show that grain sorghum-to-ethanol unit costs ($32.70 GJ À1 +36%/ À14%) are also higher than those of sweet sorghum to ethanol ($19.60 GJ À1 +11%/À64%), which is marginally lower than that of gasoline ($20.70 GJ À1 ; $0.67 L À1 gasoline). This higher cost the grain sorghum-to-ethanol process is largely due to the capital equipment costs of the  dry grind facility, unlike the sweet sorghum facility, whose capital costs are assumed fully depreciated due to being added to the existing sugarcane facility in Bella Union (Table 2). When evaluated as substitutes for gasoline, on average, sweet sorghum-to-ethanol results in negative GHG abatement cost as a result of its lower *Net N input includes total external N input and total N output. †Net N input includes both total external and internal N inputs and total N outputs. GWI and average and low unit cost relative to gasoline. The production cost of sweet sorghum to ethanol is sensitive to the electricity coproduct selling price (Table S4). When the wholesale electricity selling price is low ($0.06 kWh À1 ), the sweet sorghum pathway yields a slightly positive abatement cost. However, when electricity sells at peak price ($0.24 kWh À1 ), the facility can sell its coproduct at a higher market price, reducing the marginal production cost of the ethanol product ($6.97 GJ À1 ; $0.15 L À1 ) and achieving a low negative (cost savings) GHG abatement cost. Grain sorghum has a positive abatement cost even at the low end of the sensitivity analysis when unit costs are lowest. This is due to the accumulation of amortized capital costs and operating costs, which are not offset enough by revenues from sale of DDGS coproducts to compete with gasoline market prices ($20.70 GJ À1 ). Even at high gasoline market prices, which between 2013 and 2015 were up to $0.80 L À1 ($24.66), grain sorghum production costs are still higher and result in positive abatement costs.

Crop yields and available crop residue
Crop yields can have an important impact on system economics with production scale yield estimates often lower than in small research plot yield trials; we used average commercial field scale yields for the biorefinery in our analysis (Table 4),~70% of small plot yield trials for sorghum in Uruguay (http://www.mgap.gub.uy; http://www.inia.uy/en). The modeled grain and sweet sorghum yields are typical of the regions near Paysand u and Bella Uni on Uruguay, respectively, where these crops are grown. Farmers using better management on more productive lands could see higher yields, with improved economics and reduced GHG emissions. The frequency of pasture in grain crop rotations varies with commodity prices, farmer sense of risk, soil conservation regulations, and proximity to the biorefinery and port to reach export markets. Including  pastures in crop rotations can reduce soil erosion and N inputs and losses and increase crop yields and soil carbon (Sanderson et al., 2013;Sulc & Franzluebbers, 2014). In periods of high commodity prices, increased frequency of soybeans and grain sorghum in crop rotations near Paysand u would occur. In contrast, low commodity prices can lead to a shift toward increased pasture. Although pasture may not improve economics relative to grain even during times of low commodity prices, it potentially reduces the risk of crop failure (USDA-NRCS 2004). Consequently, the frequency of grain crops in the rotation would be expected to vary with commodity prices rather than there being an effect of increased bioenergy crop production. In addition to commodity prices, proximity to a good port for export to world grain markets also influences the frequency of grain crops and pasture in the rotation. While Paysand u has good access to a port to reach export markets, Bella Uni on is more remote so generally grain crops are more common near Paysand u, while pasture is more common near Bella Uni on. Winter double and cover crops are becoming common in Uruguay with wheat more common near Paysand u and oats and ryegrass near Bella Uni on. Because increased bioenergy crop production is not expected to affect the frequency of pasture in the rotation, while the frequency of grain commodities is expected to vary for reasons previously described, scenarios with increased pasture and grains in the rotation were evaluated while varying synchronously between the baseline and bioenergy crop rotations.

Greenhouse gas emissions and nitrogen dynamics
The identity of crops in the rotation can have a significant effect on GHG emissions, due to changes in system N with inputs/outputs and changes in crop rotation productivity affecting carbon inputs/outputs. The inclusion of pasture in the rotation had the most significant effect on N 2 O emissions both due to changes to N inputs/outputs and residual effects on subsequent crops. Although external N inputs did not appear to be correlated with N 2 O emissions, the high internal N cycling from both N mineralization and manure/urine inputs with grazing cattle in pastures did correlate well with N 2 O emissions. In spite of N outputs also being higher mainly from N consumed with forage and gaseous N emissions, net N inputs were also higher when internal cycling was included. An important point to note is that a portion of N consumed by grazing cattle is recycled as manure and urine back to the pasture rather than being removed from the system. This model behavior is consistent with observations that nitrification rates and N 2 O emissions are often more highly correlated with N turnover rates than with soil ammonium concentrations (Parton et al., 1996). In a previous Day-Cent simulation based in the USA, we observed that N 2 O emissions were correlated with N mineralization in a corn-soybean rotation both with and without winter double crops (Adler et al., 2015). N 2 O emissions varied over the crop rotation cycle due to direct effects of crops, the number of times specific crops were planted in the six-year rotation, and the legacy effects on subsequent crops. To separate the individual effects of crops on cumulative rotation N 2 O emissions, we quantified the daily N 2 O emissions during the fallow period, as well as the annual duration (Fig. 3) and the N 2 O emissions of each crop cycle in the rotations (Fig. 4). N 2 O emissions during the fallow period were higher when pasture was in the rotation due to greater intensity of N 2 O emissions, especially in the sweet sorghum rotation (Fig. 3). Greater N 2 O emissions were not due to a longer fallow period; the fallow period was longer without pasture in the grain sorghum rotation and mixed in the sweet sorghum rotation (Fig. 3). The period following pasture was most affected by its presence in the rotation. N 2 O emissions intensity increased by a factor of almost 10 with oats following pasture in the sweet sorghum rotation and~1.5 with wheat following pasture in the grain sorghum rotation (Fig. 4). The pasture effect on N 2 O emissions intensity in wheat may have been muted by it occurring twice in the rotation with the second occurrence following soybeans. The N 2 O emissions during the bioenergy crop production period was also affected by the presence of pasture due to the legacy effects, sweet sorghum in the rotation with pasture was about 1.6 times higher, while grain sorghum was only about 1.1 times higher than the rotation without pasture. Other differences in N 2 O emission intensity with specific crops were more subtle.
The accounting error, if using Tier 1 rather than Tier 3 accounting methods, was significant in this study for sweet sorghum, where the GWI from total N 2 O emissions was >30% lower with pasture when using Tier 1, but similar and only~5% lower using Tier 3 methods. This occurred due to legacy effects of legumes in the pasture being captured with Tier 3 and not with Tier 1 methods, where the sweet sorghum following pasture had a lower requirement for N fertilizer due to the 'N credit' from the pasture. Although there is a large range of uncertainty around N 2 O estimates (Del Grosso et al., 2008), this would not affect the presence of legacy N in multiple crop rotations.
The quantity and quality of carbon inputs from crop residue and roots relative to the previous rotation or the baseline rotation will affect the direction of change in soil carbon. Both with and without pasture in the rotation SOC decreased in sweet sorghum and increased in grain sorghum rotations relative to the baseline (Table 5). The direction of change in SOC was consistent with C inputs, with C inputs being lower for sweet sorghum and higher for grain sorghum rotations relative to the baseline. Garcia-Pr echac et al. (2004) observed that the soil had a significant effect on the direction of change in soil carbon with pasture in the rotation observing both increases and decreases in SOC.

Greenhouse gas emissions on a life cycle basis
Selection of a baseline rotation for scenario analysis is the most significant factor affecting results when CLCA methods are used and may be the most uncertain component of the LCA, with growing uncertainty expected in factors related to crop rotation and management to be implemented in the future. This aspect is consistent with what Plevin et al. (2014) describe regarding scenarios explored through CLCA and is also relevant to cases in ALCA where land use is a dominant factor (Soimakallio et al., 2015). Although regional factors affect the portfolio of crops grown, commodity prices have a significant effect on farmer choices and these are hard to predict from year to year, let alone over a 20-30-year timeframe commonly used in biofuel LCAs. Although economic models are commonly suggested to address this issue (Khanna & Crago, 2012), and those models are suggested as inputs into CLCAs [see Earles & Halog (2011) and Earles et al. (2012) for an example application in the forestry sector], their results are driven by underlying assumptions of commodity prices, which also host much uncertainty as a means of representing consequential change, an artifact that is also inherent in the socioeconomic complexity described by Suh & Yang (2014) and Hertwich (2014). In this study, we relied on local agricultural commodity experts to define the rotations and one would not expect economic models to significantly improve the certainty of a baseline choice.
Implications of Tier 1 and 3 GHG estimation with ALCA and CLCA methods ALCA methods are the most commonly used in GHG LCAs; however, they are not able to capture the change in GHG emissions attributable to bioenergy feedstock demand, an important prerequisite for carbon markets. This gives rise to the need to establish a 'business as usual' scenario without bioenergy feedstocks, and an 'anticipated baseline' counterfactual scenario that incorporates the increased demand for bioenergy feedstock used with CLCA methods. However, baselines required to measure additionality can be highly uncertain, especially when predicting future outcomes. We assumed that bioenergy crops did not affect the presence of pasture in the rotation; that instead rotations were influenced more by commodity prices and proximity to ports; therefore, the presence of pasture was in synchrony between the baseline and bioenergy crop rotation.
We come to the following conclusions when comparing and contrasting LCA and N 2 O estimation methods. First, within an attributional LCA framework, using IPCC Tier 1 methods to estimate N 2 O emissions for a single crop within a multiple crop rotation can result in an accounting error because it assumes that residue N cycles completely during that crop year, ignoring legacy effects of residue decomposition between years, which is especially important when the preceding crop is a legume. Legacy effects are greatest immediately following the crop having the effect, in this case pasture with legumes. Therefore, we saw a greater effect on sweet sorghum than grain sorghum when using Tier 1 LCA methods than Tier 3. Second, the choice of baseline and crop identity in rotations evaluated within the CLCA framework both affect the GWI of ethanol. Understanding the uncertainty of the baseline is important when applying CLCA methods and may be more uncertain than IPCC methods for quantifying GHG emissions. Although there is a perception that CLCA methods offer a means of capturing the environmental impacts resulting from changes or introductions of technology or policies, they cannot overcome the uncertainty that comes with modeling unknown future trajectories, including how a 'baseline' is defined, an argument raised by Plevin et al. (2014). Not only does the baseline choice affect the GWI, but so can the identity of crops in the rotation due to the legacy effects that can carry over to crops following in the rotation. Finally, ethanol produced from both grain and sweet sorghum reduced the GWI >50% relative to gasoline in >80% of the scenarios (all but three crop rotation scenarios). The U.S.'s national renewable fuel standard classifies advanced fuels, including ethanol fuel made from noncorn feedstocks, as those having a GWI that is at least 50% lower than that of gasoline. While Uruguay's alternative fuel policies aim to incentivize domestic fuel production rather than a carbon reduction target, this analysis shows climate change mitigation benefits from their investment in grain and sweet sorghum processing technology. However, production costs are high for grain sorghum and on average low for sweet sorghum, subject to electricity peak pricing, compared to gasoline. Although the Uruguayan government has specifically not implemented a subsidy (or a carbon tax on gasoline) to recover the shortfall between grain sorghum-alcohol production costs and gasoline, ethanol selling prices in Uruguay have been as high as $1.80 L À1 ($85 GJ À1 ), which allows cost recovery for both sorghum-to-ethanol pathways and in such market conditions renders each negative (cost saving) in GHG cost abatement.

Supporting Information
Additional Supporting Information may be found online in the supporting information tab for this article: