Comparison of Biofuel Life-Cycle GHG Emissions Assessment Tools: The Case Studies of Ethanol Produced From Sugarcane, Corn, and Wheat

1 Comparison of biofuel life-cycle GHG emissions assessment tools: the case
2 studies of ethanol produced from sugarcane, corn, and wheat

3
4 Pereira, L.G.1,2*, Cavalett, O.1, Bonomi, A.1,3, Zhang, Y.4, Warner, E.4, Chum, H.L.4
5
6 Brazilian Bioethanol Science and Technology Laboratory, Brazilian Center of Research in Energy and
1
7 Materials, Campinas, SP, Brazil

8
9 2
Department of Civil and Mineral Engineering, University of Toronto, Toronto, ON, Canada
10
11 3
School of Chemical Engineering, University of Campinas, Campinas, SP, Brazil
12
13 4
National Renewable Energy Laboratory, Golden, CO, USA
14
15 *Corresponding author: E-mail address: lucgpereira@gmail.com / lucasg.pereira@utoronto.ca
16
17 Abstract
18
19 The use of alternative fuels, particularly bio-based fuels, has been an important strategy to
20 achieve greenhouse gas (GHG) emission reductions compared to petroleum-based fuels.
21 However, discrepancies between results obtained by using different attributional life-cycle
22 assessment (LCA) tools have challenged the credibility of the individual assessments, and as
23 result, the progress towards or compliance with GHG mitigation targets. The objective of this
24 study was to identify the main differences and commonalities in methodological structures,
25 calculation procedures, and assumptions for the major commercial biofuel, ethanol, across
26 three public LCA tools, BioGrace (EU), GHGenius (Canada), and GREET (U.S.), and a
27 research-oriented fourth, the Virtual Sugarcane Biorefinery (VSB), a Brazilian platform for
28 sugarcane ethanol assessments.
29 The calculated emissions across models ranged from 16-45 for sugarcane, 43-62 for corn,
30 and 45-68 g CO2eq MJ-1 for wheat ethanol. Harmonizing the three public models with VSB
31 assumptions for sugarcane ethanol produced in Brazil, the range was reduced to 16-17 g
32 CO2eq MJ-1 for sugarcane ethanol. Agricultural production (e.g., N2O emissions from
33 fertilizers; energy and fuel use; straw field-burning; and limestone application) and ethanol
34 shipping were found to be the major causes for variations for differences calculated for
35 sugarcane ethanol. Similarly, harmonizing BioGrace and GHGenius calculations using
36 GREET assumptions for U.S. corn ethanol generated nearly identical results (models varied
37 within a 3% range). The coproduct treatment method was found to be the most influential
38 parameter in the variations calculated for both corn and wheat ethanol. The study provides
39 recommendations for promoting transparency in LCA calculations and assumptions across
40 the tools used in research and development of biofuels or for regulatory tools. The application
41 of the tools as part of GHG emissions accounting requirements is often defined via
42 regulations and differences and/or conflicting assumptions set-forth in these models lead to
43 most differences observed.
44
45 Keywords: LCA, GREET, GHGenius, BioGrace, greenhouse gas emissions, commercial
46 ethanol, harmonization
47
48
Pursuant to the DOE Public Access Plan, this document represents the authors' peer-reviewed, accepted manuscript.
The published version of the article is available from the relevant publisher.
49 Abbreviations
50 CARB: California Air Resources Board
51 CCLUB: Carbon Calculator for Land Use Change from Biofuels Production, sub-model of
52 GREET
53 CH4: methane
54 CNPEM: Brazilian Center for Research in Energy and Materials
55 CO2: carbon dioxide
56 CO2eq: carbon dioxide equivalent
57 CONCAWE: Conservation of Clean Air and Water in Europe
58 CRC: Coordinating Research Council, Inc.
59 CTBE: Brazilian Bioethanol Science and Technology Laboratory
60 EPA: U.S. Environmental Protection Agency
61 EU: European Union
62 EU-CAR: European Council for Automotive Research and Development
63 EU-RED: European Union Renewable - Energy Directive
64 FQD: Fuel Quality Directive, EU
65 GHG: greenhouse gas
66 GREET: Greenhouse Gases, Regulated Emissions, and Energy Use in Transportation, by
67 Argonne National Laboratory, ANL, U.S.
68 GWP: global warming potential – 100-year
69 IPCC: Intergovernmental Panel on Climate Change
70 ISO: specific standard from the International Organization of Standards
71 JEC: Joint European Commission
72 km: kilometer
73 LCA: life-cycle assessment (GHGenius); life-cycle analysis (GREET)
74 LCFS: Low Carbon Fuel Standard
75 MJ: megajoule
76 N2O: nitrous oxide
77 RFS2: U.S. Renewable Fuel Standard 2
78 RSB: Roundtable on Sustainable Biomaterials
79 UK-RTFO: United Kingdom Renewable Transportation Fuels Obligation
80 VSB: Virtual Sugarcane Biorefinery, CBTE, CNPEM, Brazil
81 WTT: well-to-tank
82 WTW: well-to-wheel
83
84
85
86
87
88
89
90
91
92 1. Background
93 The use of bio-based alternative fuels has been considered an important strategy to achieve
94 reductions in greenhouse gas (GHG) emissions from the transportation sector dominated by
95 petroleum fuels, promote rural economy and improve energy security. Government agencies
96 around the world have developed initiatives and policies to encourage the production and use
97 of biofuels, which could contribute to multiple sustainability goals. Increasingly the growing
98 variety of biofuels, including high energy density hydrocarbon fuels, makes the issue of
99 reducing GHG emissions of increased importance [1–6].
100 To measure the progress towards one of the primary goals of biofuel policies, i.e., GHG
101 emission reductions, models have been developed to quantify life-cycle GHG emissions of
102 biofuels and their reference fuels (typically petroleum gasoline, diesel, and jet fuels). Some
103 models were designed to comply with regulatory requirements, whereas others were adopted
104 and/or modified from existing research and development tools investigating multiple
105 pathways for fuels coupled to vehicle life-cycle systems.
106 This study examined three publicly available life-cycle GHG emissions models for
107 transportation fuels: GHGenius [7,8] used in Canadian provinces to determine the carbon
108 intensity of fuels under the Low Carbon Fuel Standards (LCFS); GREET (Greenhouse Gases,
109 Regulated Emissions, and Energy use in Transportation Model) [9], used by the U.S.
110 Environmental Protection Agency (EPA) along with other models for the rulemaking of the
111 Renewable Fuel Standard 2 (RFS2). Derived GREET versions are used by the California Air
112 Resources Board (CARB) LCFS (U.S.); and BioGrace (BIOfuel GReenhouse gas emissions
113 Alignment of Calculations in Europe) [10], simplified for use in compliance with the
114 European Union Renewable Energy Directive (EU-RED) and the Fuel Quality Directive
115 (FQD), with harmonized data for EU nations.
116 Because the LCA models were designed for different purposes, significant variations have
117 been observed in GHG emissions estimates, especially for biofuels [11–22]. Many factors,
118 including assumptions, input data, treatment of coproducts, level of details built in the
119 calculation structure, and specific attributes of the Life-cycle Assessment (LCA) approach
120 utilized could influence the estimates of carbon intensity per functional unit. In addition,
121 much of the literature does not identify the specific model version used nor the level of
122 maturity of the technology, making meaningful comparisons of results difficult, and, at times
123 misleading. The inconsistency among these GHG modeling tools and their use has led to a
124 great variability in GHG emissions results for the same fuel pathway such as ethanol from
125 sugar or starch crops, biodiesel from soybean or rapeseed which are commercial and bagasse
126 or corn stover to ethanol which are in demonstration or initial commercialization. Macedo et
127 al. [6] pointed out technical challenges in the evaluation of biofuels, including the need for
128 reliable data (agronomic and conversion) and for a higher level of scientific consensus on a
129 series of aspects, such as the treatment of coproducts, land use change, and reference systems
130 [11,12,15,16].
131 One of the most debated pathways is ethanol produced from corn grown in the U.S.: some
132 studies concluded that its GHG emissions would be nearly twice as high as those from
133 gasoline [16,17], mainly driven by the (direct and indirect) land use changes induced by
134 diverting corn to produce ethanol. Other studies found that corn ethanol offers advantageous
135 GHG reductions of up to 40% in comparison with gasoline in scenarios which maintain the
136 reference land use as continuing to be productive or improved agricultural land [11,18–20].
137 In the case of Brazilian sugarcane ethanol, Cavalett et al. [21], Seabra et al. [22], Wang et al.
138 [13] reported GHG emissions savings between 67% and 86% relative to the baseline
139 gasoline. Khatiwada et al. [14] assessed GHG emissions for Brazilian sugarcane ethanol
4
140 under North American (i.e., EPA’s RSF2 and California Air Resources Board CARB LCFS)
141 and European Directives and regulatory schemes (i.e., EU-RED and United Kingdom
142 Renewable Transportation Fuels Obligation (UK-RTFO)). The authors found that GHG
143 emissions savings at that time could vary as much as 30% within the same method (e.g., the
144 EPA RFS2 approach can range between 61% and 91%) and more than 40% using the
145 methods required by different regulations (e.g., CARB LCFS approach 31% and EU-RED
146 approach showed 72%). Agricultural practices (especially soil carbon and nitrogen
147 dynamics), coproduct credits from surplus electricity, and uncertainties around economic
148 modeling approaches for indirect land use change were the major drivers of methodological
149 divergences.
150 The discrepancy in GHG emissions estimates from the application of different LCA models
151 noted in the previous analyses affected the acceptance of LCA results and their use in the
152 biofuel policy context and the reported progress towards meeting the established GHG
153 emissions reduction targets. A few studies on the comparison of LCA models used for
154 regulatory purposes investigated methodological differences. Hennecke et al. [23] compared
155 BioGrace versus the RSB (Roundtable on Sustainable Biomaterials) version approved under
156 the EU-RED. Chum and Warner [24] performed a harmonization study for dry-mill corn
157 ethanol production based on natural gas energy source using GHGenius and BioGrace
158 models with GREET 2015 data as model inputs. Harmonized model results agreed well with
159 each other within 2.5%. O’Connor [25] and Unnasch et al. [26] evaluated a broader range of
160 LCA models for the Coordinating Research Council, Inc. (CRC) of petroleum and
161 automotive equipment.
162 In this paper, in addition to the three publicly available LCA models the Virtual Sugarcane
163 Biorefinery (VSB) [27], developed by the Brazilian Bioethanol Science and Technology
5
164 Laboratory (CTBE) served as reference for the sugarcane ethanol production in Brazil and
165 its parameters served to harmonize the three public models. The primary objective of this
166 study was to quantify the main differences and identify commonalities in methodological
167 structures, calculation procedures, and assumptions, to understand how and why the models
168 generate different GHG emissions results. In addition, this comparative analysis provides
169 recommendations for more consistent and transparent LCA methodologies and best practices
170 reporting results of GHG emissions for biofuel pathways used globally.
171
172 2. Method
173 2.1. Life cycle assessment
174 The LCA methodology is often used to assess the environmental impacts associated with a
175 product, process or activity, by the identification and quantification of energy and materials
176 flows used, as well as waste and emissions released. The approach has been utilized as a
177 standard to estimate life-cycle GHG emissions of biofuels. The general framework for
178 conducting the assessment can be found in the International Organization for Standardization
179 (ISO) documents [28,29]. The general procedure includes the following steps: defining
180 scope, system boundaries, functional unit, and reference systems; determining mass and
181 energy flows; and treating coproducts; and assigning impacts to energy and material flows.
182 Figure 1 shows a schematic overview of the typical life cycle (well-to-wheel or WTW) of
183 ethanol produced from biomass. The scope of the LCA approach usually takes into
184 consideration the agricultural (including cropping, harvesting, and other relevant operations),
185 transport, industrial (conversion of feedstock to biofuel), transport of the fuel through
186 distribution, and fuel use stages. Inputs and outputs from each stage may vary depending on
187 the type of feedstock considered, any preprocessing conducted, type of conversion plant, the
6
188 location of these different activities, the considered system boundaries, and the LCA
189 assumptions utilized.
190
191
192 Figure 1. Schematic overview of the life cycle of ethanol fuel and the main inputs and
193 emissions from each stage
194
195 2.2. Considered LCA models
196 In this section, the LCA models utilized for the comparison carried out in this paper are
197 briefly described. Detailed assumptions and calculation mechanisms within the models are
198 presented in section 3.
199
200 2.2.1. GREET
201 The GREET model [9] was developed in 1996 by the Argonne National Laboratory (ANL)
202 for the U.S. Department of Energy and sponsored by the U.S. Department of Energy's Office
203 of Energy Efficiency and Renewable Energy. It was originally designed to be a simple-to-
204 use tool through which researchers could evaluate fuel-cycle energy and emission impacts of
205 various transportation technologies. Since its inception, the model continues to be updated
206 and expanded by the ANL. The most recent versions are the GREET1 2017 for fuel-cycle
207 analysis, GREET2 2017 for vehicle-cycle analysis, and a graphic interface named
208 GREET.net 2017. The model includes more than 100 fuel pathways such as petroleum fuels,
209 natural gas fuels, biofuels, hydrogen, and electricity produced from various energy feedstock
210 sources. Three classes of vehicles and various technologies are addressed. GREET version
211 2016 was used in this study.
212 In 2007, GREET version 1.8b was modified by Lifecycle Associates for CARB to quantify
213 GHG emissions for implementing LCFS in California, where it is used for regulatory
214 purposes and updated periodically.
215
216 2.2.2. GHGenius
217 GHGenius [7][8] is a model developed for Natural Resources Canada by (S&T)2 Consultants
218 based on the 1998 version of Lifecycle Emissions Model by Delucchi [30]. The model
219 calculates energy and emissions associated with conventional and alternative fuel production
220 starting with the input from the past years (beginning with 1995) and uses projections into
221 the future (2050). This study made use of GHGenius version 4.03 of 2013, which is capable
222 of modeling multiple geographic regions, including not only country-level analyses for
223 Canada, U.S., Mexico, and India, but also accounting for regional specificities. The public
224 model version 5.0a is currently used as the basis for regulations; it was released in 2018, after
225 this study was completed.
226
227 2.2.3. BioGrace
228 BioGrace [10] is a spreadsheet model used for the calculation of biofuel GHG emissions
229 originated from the European cooperative harmonization effort, in which country model
230 owners from Germany, the Netherlands, Spain, and the United Kingdom developed the
231 calculator to implement the EU-RED and the FQD. The current model manager is the
8
232 Institute for Energy and Environmental Research (IFEU) from Germany (BioGrace 4d
233 version was used).
234 The calculation is based on a database with default values (EU averages) of 22 commercial
235 feedstock/biofuels pathways elaborated by a joint group of experts, including the Joint
236 European Commission (JEC) from the Joint Research Centre (JRC), the European Council
237 for Automotive Research and Development (EUCAR), and the Conservation of Clean Air
238 and Water in Europe (CONCAWE). The complete and updated database and their models
239 are not public, although results from multiple fuels are published periodically. BioGrace 4d
240 uses a set of default values from 2011 that is accessible to users [31]. The model also allows
241 users to input their own data as an approved voluntary scheme to demonstrate compliance
242 with the EU-RED.
243
244 2.2.4. Virtual Sugarcane Biorefinery
245 The VSB was developed by the Sustainability Division of the CTBE from the Brazilian
246 Center for Research in Energy and Materials. The tool evaluates sugarcane biorefinery
247 configurations, combining chemical engineering computer simulation software (e.g., Aspen
248 Plus), mathematical modeling, and sustainability assessment methods with the objective of
249 identifying and evaluating technical parameters and sustainability impacts related to novel
250 and existing biorefinery configurations. Sustainability impacts calculated in this tool include
251 environmental (e.g., climate change, human toxicity, ecotoxicity, particulates, acidification,
252 eutrophication, energy, land, and water use), economic (e.g. internal rate of return, net present
253 value, and levelized production cost) and social (e.g. number of jobs, occupational accidents,
254 wages, educational level, and gender distribution) aspects [27,32–47].
255
256 2.3. Approach for comparison of the LCA models
257 All four models were used to estimate GHG emissions impacts for sugarcane ethanol
258 produced in Brazil. GREET and GHGenius were utilized for assessing the impact of corn
259 ethanol produced in the U.S., whereas BioGrace calculated the impact of corn ethanol
260 produced in the EU. GHGenius and BioGrace were used to assess wheat ethanol production,
261 the first with data from the U.S. and the latter from the EU. For the sake of consistency,
262 modeling was performed for the same year, when possible. For GREET and GHGenius, the
263 year was set to 2015; VSB already utilizes average values for sugarcane ethanol in 2015 as
264 default. The BioGrace 4d version, however, does not support changes to the default model
265 year (2011).
266
267
268 Figure 2. Schematic representation of the main methodological steps for the comparison of
269 LCA models.
270
271 Figure 2 depicts the methodological steps used to compare LCA models. Step (1) of the
272 analysis compares the main characteristics, parameters, and assumptions of the models. Step
273 (2) extracts the life-cycle inventories from the models and converts to common units (e.g.,
10
274 mass and energy per tonne of feedstock, kilometers (km) of transportation). This step
275 generates preliminary numerical comparison among models which use their own inputs and
276 categories in different units, thus making direct comparison more difficult. Step (3)
277 harmonizes activity categories (e.g., fertilizer manufacture, diesel use, and natural gas use)
278 associated with the life cycle of ethanol production and use. The categories were harmonized
279 across the models when the values were extracted from each model for each category. For
280 instance, GHGenius presents a category named “land use changes, cultivation”, which
281 aggregates a series of items such as limestone application, N fertilizer use, and field-burning
282 of sugarcane straw with manual harvest, among others. This definition of land use changes
283 associated with cultivation used in GHGenius is different than that in the other models. The
284 harmonization process, in this case, was not straightforward; the intricate calculation
285 mechanisms of the models had to be analyzed for each specific case. The comparison of the
286 actual results calculated for the GHG emissions of ethanol followed using each LCA model.
287 This part of the analysis was intrinsically linked to the other steps, given that the differences
288 observed in the calculated results were discussed based on information obtained in the
289 previous steps. In Step (4), a harmonization procedure utilizing modified steps to select
290 parameters and assumptions within the models was performed using the model’s default
291 values as starting points. The objective was to assess to what extent a specific step would
292 affect the results and how close final figures obtained would be in relation to each other.
293 In summary, the modification steps described for sugarcane ethanol in Brazil and corn
294 ethanol in the U.S. accomplished the following: (1) applying consistent approaches for
295 coproducts treatment (i.e., economic allocation for sugarcane ethanol and substitution for
296 corn ethanol); (2) removing overseas transportation (for sugarcane ethanol only), equaling
297 and harmonizing ground transportation distances and GHG emissions from ethanol used in
11
298 vehicles; (3) harmonizing nitrogen, limestone (for both ethanol cases), and field-straw
299 burning GHG emissions (for sugarcane ethanol only); (4) harmonizing energy use in
300 agricultural operations; and (5) harmonizing GHG emissions from the ethanol production
301 processes (industrial conversion stage) and yields to those of VSB for sugarcane ethanol in
302 Brazil and of GREET for corn ethanol in the U.S.
303
304 3. Results and discussion
305 3.1. Comparison of the models
306 A comparison of the LCA models was performed based on selected characteristics
307 summarized in Table 1. In this section, key drivers of differences among modeling results
308 are discussed (e.g., LCA approach and treatment of coproducts, upstream life-cycle data, and
309 N2O emissions).
310 Although GREET and GHGenius are used by policymakers for regulatory compliance in the
311 U.S. and Canada, respectively, BioGrace is the only model developed to assess compliance
312 of biofuels in the EU. Modified versions of GREET are used for compliance in the CARB
313 LCFS.
12
314
315 The LCA scope is consistent mostly among GREET, GHGenius, and VSB. BioGrace,
316 however, it is a well-to-tank (WTT) calculation following the EU-RED legislation specified
317 values of various parameters. The model includes a threshold value or cut-off criterion of 0.1
318 g CO2eq per MJ of fuel for inclusion of component inputs or processes in the accounting of
319 the total GHG emissions of the biofuel pathway. The JEC model is capable of modeling both
320 WTT and tank-to-wheel emissions, making use of specific values for the chosen propulsion
321 system [31]. GREET and GHGenius have coupling choices of WTT with several types of
322 vehicle cycle modeling tools to obtain overall WTW results in public models. VSB also
323 accounts for the vehicle emissions and provides multiple functional units such as CO2eq
324 emissions in terms of km traveled and include the exhaust emissions from the use of fuel in
325 vehicles, as well as the emissions from the whole supply chain.
326 All four models provide life-cycle GHG emissions inventories; however, only VSB applies
327 a life cycle impact assessment method and assesses other local and global environmental
13
328 impacts. Life-cycle GHG emissions associated with gasoline (in g CO2eq per MJ of fuel)
329 used as a reference or baseline to estimate the GHG emissions reductions obtained by
330 biofuels can differ as much as 12% across models (95.0 g CO2eq per MJ for GHGenius and
331 83.8 g CO2eq per MJ for BioGrace). The 12% difference includes producing, transporting
332 and refining the various crude oils and alternative sources (tar sands or shale oil) used to
333 produce the fossil fuel in the countries or states that are using the fossil and renewable fuels.
334 Climate impact factors are typically expressed in terms of global warming potentials using a
335 time horizon of 100 years (GWP100), although the GWP values are updated over time with
336 improvements from climate science. BioGrace utilizes GWP100 from the IPCC 2001 report
337 [48] as default [10]. It provides, however, the option of switching the factors to those from
338 the IPCC 2007 report [49]. GREET and VSB use GWP100 from the IPCC 2013 report [50] as
339 default. GREET allows users to select values among all five IPCC reports [48–52].
340 GHGenius uses factors from the IPCC 2007 [49] report as default; however enabling the use
341 of factors from IPCC 1995 and 2001 reports [48,52]. For the sake of comparison,
342 characterization factors for all models were harmonized to use GWP100 values in the IPCC
343 2007 report [49].
344 GREET allows for the inclusion of GHG emissions associated with direct land use change
345 due to corn crop management and soil carbon data from the CENTURY model, Carbon
346 Online Estimator (Tier 3) for the U.S. and induced land use change based on its Carbon
347 Calculator for Land Use Change from Biofuels Production (CCLUB) module [53]. The
348 CCLUB model combines land conversion (area and type) data from Purdue University’s
349 Global Trade Analysis Project model and the CENTURY model for estimates of U.S. soil
350 carbon and Winrock International for international land use [54–57].
351
14
352 3.1.1. LCA approach and treatment of coproducts
353 All four models use an attributional LCA approach as default, meaning that the focus of the
354 models is on describing the environmentally relevant physical flows (e.g., materials and
355 energy) to and from the environment for the biofuel production at defined system boundaries
356 [58–62]. This differs from a consequential LCA approach, which aims to describe how
357 environmentally relevant flows will change in response to possible substitution of products’
358 decisions [59].
359 The choice of the procedure used to treat coproducts is one of the most controversial topics
360 in LCA. The issue arises when a system produces more than one valuable output, as in a
361 multi-functional system. The concern is associated with biofuels production systems because
362 other useful products are often coproduced with the fuel of interest. The ISO 14040 and
363 14044 documents recommend avoiding allocation whenever possible either through
364 subdivision of processes or by expanding the system boundaries to include the functions
365 associated with the coproducts generated. This includes displacement or substitution methods
366 that are intrinsically more complex as their implementation requires significant amount of
367 market data in laborious levels of detail [61]. For instance, in the case of corn ethanol, a major
368 coproduct is dried distillers’ grains with soluble materials (DDGS), a protein rich animal
369 feed, for beef, dairy, swine, and poultry. DGS (wet form) is also sold to closely located beef
370 and dairy animals farms. DDGS substitutes corn, soy meal, and urea at specific displacement
371 ratios determined, which is one example of the expansion of the boundaries. The aggregate
372 amounts of the various displaced conventional products would have generated GHG
373 emissions that are offset by the displacement ratios and amounts of the corresponding nonfuel
374 product output. The emissions displaced are the credits for production of fuel products from
375 the total system GHG emissions [61]. This example also includes another level of expansion
15
376 of the system given by the amount of methane generation credit for avoided methane in dairy
377 production. Current dry mills (70%) also extract corn oil (non-edible) for the production of
378 biodiesel and thus the boundaries are further increased with this additional energy coproduct.
379 In summary, multi-functionality here is treated by subtracting from the product system the
380 impacts from reference products (in the example corn, soy meal and urea) that may be
381 displaced by the coproducts in the market. The substitution method can be considered a step
382 into the consequential LCA approach.
383 When it is not possible to avoid allocation, the main recommendation by ISO is to use
384 methods that reflect the physical properties or other relevant variables [60]. The most recent
385 ISO document on this matter is the Sustainability Criteria for Bioenergy [63], which provides
386 further guidance on the treatment of coproducts. Documentation of the selected procedures
387 for coproducts treatment is needed and justified, including sensitivity analysis to illustrate
388 the consequences of using alternative procedures.
389 The models investigated in this study made use of different coproduct treatment approaches
390 as default: the VSB considers economic allocation while BioGrace makes use of energy for
391 partitioning as per regulation, although JEC [31] suggests substitution as the most appropriate
392 approach. GHGenius utilizes the substitution method, whereas GREET allows the user to
393 calculate multiple approaches depending on the biofuel pathway. For instance, the impacts
394 of sugarcane ethanol and surplus electricity produced in Brazil are allocated according to
395 energy as default, whereas the impacts of corn ethanol in the U.S. can be calculated by
396 substitution or other allocation methods.
397
398 3.1.2. Upstream life-cycle data
16
399 The life-cycle stages of a process or product include extraction, manufacture, logistics, and
400 use; a product used as an input to a process carries the impact load from the previous stages.
401 To account for these impacts in the assessment, GREET and GHGenius use internal
402 calculation mechanisms based on sector inventories: power, transportation, industrial, and
403 resource extraction. BioGrace and VSB utilize the JRC-EUCAR-CONCAWE database v4.a
404 [31] and the Ecoinvent database v2.2 [64], respectively.
405 Main relevant items in terms of GHG impacts include fertilizers, diesel, and natural gas,
406 among others, accounting for their manufacture and use. Table 2 presents the upstream life-
407 cycle data for selected inputs. The impacts can vary significantly across the models. For
408 instance, regarding diesel, the impact can vary as much as 30% with 116.4 g CO2eq per MJ
409 estimated by GHGenius (with 75% of the total impact associated with the combustion stage)
410 and 81.6 g CO2eq per MJ by the VSB (with 55% of the total impact associated with the
411 combustion stage); for the nitrogen fertilizer manufacture, variation can be as much as 43%
412 with 5.88 g CO2eq per kg of nitrogen for BioGrace and 3.35 g CO2eq per kg of nitrogen for
413 the VSB.
414
17
415 3.1.3. Nitrous oxide (N2O) emissions
416 The contribution of N2O is an important variable in the calculation of climate impacts,
417 magnified by its high characterization factor that can be as much as 300 times greater than
418 that of CO2 depending on the IPCC GWP method chosen. N2O emissions on biofuel
419 pathways come mainly from nitrogen fertilizer application and organic matter decomposition
420 [65], depending on soil type, climate, crop, tillage method, and fertilizer and agricultural
421 residues application rates.
422 LCA studies accounting for N2O emissions often utilize the default emissions factors
423 published by IPCC, which present estimates from several sources [66]: direct soil emissions
424 of N as N2O, at 1% of synthetic N fertilizer application and 2% for manure; volatilization of
425 N as NH3 at a rate of 10% of total N in the case of synthetic N application or 20% of total N
426 in the case of manure application, with 1% of the N in the NH3 converted to N2O; and runoff
427 and leaching to groundwater as nitrate at 30% of total N applied, with 0.75% of it converted
428 to N2O. The total resulting effect is that 1.325% of N in synthetic fertilizer is emitted as N2O
429 (due to direct emissions of N2O plus indirect N2O emissions from the conversion of emissions
430 of NH3 and nitrate to N2O).
431 Table 3 presents the emissions factors for direct and indirect N2O emissions from fertilizers
432 and agricultural residues assumed by each model. Although the GHG models investigated in
433 this study make use of the IPCC 2006 method as a basis to account for field emissions, small
434 differences in the assumptions may lead to significant variability in the obtained results.
435 For instance, regarding direct N2O emissions associated with the use of N-fertilizer,
436 BioGrace and the VSB use the default IPCC values for the direct N2O emissions (equivalent
437 to 1%), whereas the others consider different values for the crops (GHGenius uses 1% for
18
438 wheat and 1.25% for corn and sugarcane, whereas GREET assumes 0.895% for Brazilian
439 sugarcane and 0.9% for corn).
440
441
442 3.2. Life-cycle inventories
443 Tables 4 and 5 present the main agricultural and industrial inputs and products logistics and
444 transportation parameters for ethanol from sugarcane, corn, and wheat, respectively,
445 considered in each LCA model. Detailed lists of all the agricultural and industrial inputs used
446 for ethanol produced from the three types of biomass can be found in Tables 1S, 2S, and 3S
447 of the supplementary material.
19
448
449
450 3.2.1. Sugarcane ethanol production in Brazil
20
451 The differences observed for most of the agricultural and industrial inputs reflect the sources
452 of data selected: GHGenius uses average values from three studies [22,67,68] with the
453 exception of diesel used in agricultural machinery. GREET uses data documented in Wang
454 et al. [19] including the same references as GHGenius, and BioGrace derives data from
455 Macedo et al. [68] in 2008, whereas the data for VSB are based on recent agriculture and
456 industrial sectorial data, recommendations from experts, literature, and process simulations
457 [27].
458 A key driver, the application of limestone, triggers CO2 emissions due to its degradation:
459 GHGenius assumes an application rate of more than double the values used by the other
460 models. GHGenius uses the value reported by Macedo et al. [68] in the estimate, which
461 considers an application rate of 1,900 kg CaO per ha. In comparison, Seabra et al. [22] and
462 Macedo et al. [67] reported a field application rate of 450 kg and 366 kg CaO per ha,
463 respectively, in 2011 and 2008, from agronomic best practices at these times, and the 2011
464 publication projected rates to 2020.
465 Another important parameter is the proportion of mechanized harvested sugarcane versus the
466 manual harvesting of field-burned straw, an old practice of the ethanol and sugar industry.
467 However, since 2002, a series of state and federal laws have gradually reduced sugarcane
468 field burning practices in face of concerns with climate change and health effects [69]. The
469 models investigated in this study give the user a choice of the percentage of manual sugarcane
470 harvesting. Changing the default values can be tricky and complex, given that models are not
471 designed for non-LCA designers as users. Original default values of the models were
472 maintained for this comparative assessment prior to harmonization. GREET assumes a time
473 series for straw burning with values ranging from 95% of manual harvesting in 1995 to 14%
474 in 2015. The value for 2015, which was used in this study, is close to the default value of
21
475 VSB tool of 18.4%. BioGrace, on the other hand, considers that 100% of the straw is field-
476 burned as default. GHGenius assumes no field burning as default; this value, however, may
477 be manually set by the user.
478 All four models assume that sugarcane ethanol is produced in Brazil. VSB uses domestic
479 transportation by truck from the producer through the distribution system to tank whereas the
480 other models include not only Brazilian ground distribution but also overseas shipping from
481 Brazil to the destination country.
482
483 3.2.2. Corn ethanol
484 The GREET model uses as default weighted-average values of the three types of corn mills:
485 dry mills without corn oil extraction (17.7%); dry mills with corn oil extraction (70.9%); and
486 wet mills (11.4%) [18]. Therefore, the values presented in Table 5 for the industrial stage in
487 the GREET model are, in fact, of the amounts of inputs and outputs used by the different
488 types of existing mills. Detailed industrial inputs for ethanol produced from the three types
489 of mills can be found in Table 4S of the supplementary material. In terms of agricultural
490 inputs, GREET makes use of the study by Wang et al. [70] as a basis to project values for
491 2014. The same values are used in the following years. Other configurations can be selected.
492 By using GHGenius, it was possible to model an average U.S. dry-mill corn ethanol facility
493 in 2015 by selecting the U.S. as the modeling location. In GHGenius, the base year for which
494 default data for corn agricultural production and the amount of inputs was originally
495 developed is 1994. The model adjusts these inputs for 2015 according to trends from 1964 to
496 2011 [71].
497 BioGrace models an average EU dry-mill corn ethanol facility powered by a natural gas
498 combined heat and power (CHP) system using the default 2008/2009 dataset derived from
22
499 Neeft [10]. Table 5 compares the primary model parameters used in the three models for
500 corn and ethanol production and transportation.
501
502 3.2.3. Wheat ethanol
503 Only GHGenius and BioGrace include a wheat ethanol pathway. The GHGenius model
504 assumes an average-wheat ethanol facility using natural gas CHP for steam and power
505 production, with on-site electricity generation. Similar to the corn ethanol pathway, the base
506 year for wheat agricultural production and the amount of inputs used in the U.S. is 1994
507 extrapolated to 2015. BioGrace models five types of wheat ethanol plants: four with on-site
508 electricity generation (unspecified process fuel, steam from lignite CHP, steam from straw
509 CHP, and steam from natural gas CHP) and one with electricity consumption from the grid
510 (steam from natural gas boiler). For comparative reasons, steam from CHP was selected for
511 evaluation in this study using the default 2008/2009 dataset based on Neeft [10]. Data for
512 wheat and ethanol production is summarized in Table 3S. The different locations in part
513 explain the variations in model inputs, such as fertilizer application rates and electricity
514 usage.
515
516 3.3. Comparison of calculated GHG emissions
517 Differences in net life-cycle GHG emissions impacts across the models are shown in Figure
518 3. Impacts values for the various categories are found in Table 5S of the supplementary
519 material.
23
520
521
522 Figure 3. Greenhouse gas emissions impacts of ethanol produced from sugarcane, corn, and
523 wheat in g CO2eq per MJ of ethanol calculated by GREET, GHGenius, BioGrace, and VSB
524 models.
525
526 Brazilian sugarcane ethanol estimated life-cycle GHG emissions vary as much as 65% across
527 models, with GHGenius presenting the highest value at 45.1 g CO2eq per MJ of ethanol
528 compared to 16-24 g CO2eq per MJ obtained with the other models. For GHGenius, the main
24
529 differences observed in comparison to the other models result from data on fuel and energy
530 use, limestone, and fertilizer GHG emissions at the agricultural stage (farming), as well as
531 low energy efficiency bagasse combustion at the industrial stage. GHG emissions from the
532 use of diesel are responsible for 7.8 g CO2eq per MJ of ethanol according to GHGenius, while
533 GREET, BioGrace, and VSB estimate 4.3, 1.3, and 2.2 g CO2eq per MJ of ethanol,
534 respectively. As expected, the GHGenius model results for limestone application are 4.9 g
535 CO2eq per MJ of ethanol, which is 12 times compared to BioGrace (0.4 g CO2eq per MJ),
536 with GREET (1.3 g CO2eq per MJ) and VSB (0.9 g CO2eq per MJ) having intermediate
537 values. Fertilizer and soil N2O emissions are estimated at 10.8 g CO2eq per MJ of ethanol by
538 GHGenius in comparison to 1.7 by GREET, 3.3 by BioGrace, and 7.9 g by VSB (in g CO2eq
539 per MJ). GHGenius also includes GHG emissions from sugarcane bagasse combustion (5.1
540 g CO2eq per MJ of ethanol), whereas the other models consider carbon emissions from
541 bagasse combustion as biogenic and, therefore, do not include them in the life-cycle results;
542 GREET calculates them.
543 The amount of diesel used in machinery operation (in L per tonne of sugarcane) was
544 estimated as 2.9 by GHGenius, which was higher than the 1.1 by GREET or 0.8 by BioGrace,
545 and 1.9 by VSB, as shown in Table 4. In addition to the amount used, GHGenius estimated
546 a higher life-cycle GHG emissions factor for diesel (production and combustion in
547 agricultural machinery) when compared to GREET, BioGrace, and VSB: 116.4 in
548 GHGenius, 90.2 in GREET, 87.6 in BioGrace, and 81.6 in VSB, all in g CO2eq per MJ of
549 diesel fuel, as shown in Table 3. In the case of limestone, a similar explanation is valid. The
550 discrepancy observed for the impact calculated by GHGenius in comparison to the other
551 models may be attributable to two main factors, such as the higher application rate of lime
25
552 (Table 4) and the much larger upstream life-cycle impact value associated with limestone
553 (Table 3).
554 Ethanol shipping is also a very important category for sugarcane ethanol; values calculated
555 through GREET, BioGrace, and VSB would be similar with a maximum variation of 15%:
556 (in g CO2eq per MJ of ethanol) 16.8 for GREET, 18.9 for BioGrace, and 16.1 for VSB.
557 The influence of the approach for the treatment of coproducts utilized by the models when
558 evaluating sugarcane ethanol is minor, as shown in Table 6S of the supplementary material.
559 Excess electricity generated in the production of sugarcane ethanol corresponds to 5% of the
560 total impact in GREET (energy allocation), 0% in BioGrace (energy allocation), 9% in
561 GHGenius (substitution method), and 4% in VSB (economic allocation). Results from the
562 impact of excess electricity production in this paper are specifically linked to the default
563 assumptions considered in the investigated models. Khatiwada et al. [14] stated that, if the
564 system expansion method (substitution method) were adopted, the credits associated with
565 electricity generated could even offset all the GHG emissions resulting from ethanol
566 production in Brazil. However, the calculation of such credits depends on the selection of
567 several factors and assumptions (e.g. average electricity generation by biorefineries, type of
568 biomass used as feedstock (sugarcane bagasse and/or straw), electricity mix considered, and
569 GHG emissions related to the sources of electricity considered), which are associated with
570 uncertainties and variabilities, and whether it substitutes base or marginal electricity
571 production. In addition to the electricity generated as a coproduct, the majority of sugarcane
572 biorefineries in Brazil coproduce ethanol and sugar [32] – considering this fact could also
573 lead to variability in the partition of the GHG emissions impacts obtained. The VSB model
574 can account for variations in the portfolio of the products of a sugarcane biorefinery, as well
26
575 as for the use of additional feedstocks for electricity production (in field mechanized
576 sugarcane harvest) [39].
577 For corn ethanol, life-cycle GHG emissions obtained vary as much as 30% across models,
578 with BioGrace estimating 43.4 (specific process), GHGenius 61.9 (average) and GREET
579 57.7 (country average) g CO2eq per MJ of ethanol, respectively. The main differences
580 observed in this case are due to the treatment of the coproducts utilized. The default method
581 used by BioGrace (energy) led to a 50% partitioning of GHG emissions between ethanol and
582 its coproducts, whereas the substitution method used by GREET and GHGenius provides a
583 credit of 12.8 and 16.7 g CO2eq per MJ, respectively, for non-energy products to ethanol
584 (Figure 3 and Table 5S). GHGenius estimates GHG emissions from fertilizers at 20.2 g
585 CO2eq per MJ of ethanol in comparison to 16.2 g by GREET and 4.3 g by BioGrace. The use
586 of off-site electricity, coal, and other inputs required for corn ethanol production resulted in
587 a total of 18.1 g CO2eq per MJ of ethanol by GREET, 15.9 g by GHGenius, and 0 g by
588 BioGrace (electricity from a natural gas steam plant generates corn ethanol power needs
589 without crediting excess electricity produced). Table 5 shows the differences in fertilizers,
590 limestone, and diesel inputs at the agricultural stage, as well as the demand for electricity and
591 fuels in the industrial stage between BioGrace and the other two models.
592 The same observations made for corn are valid for wheat ethanol: BioGrace presented the
593 lowest GHG emissions impacts due to the treatment of the coproducts utilized. Additionally,
594 the discrepancy presented in terms of the inputs considered by BioGrace and GHGenius, in
595 this case, may be explained by the fact that BioGrace contemplates the production of wheat
596 in the EU, whereas GHGenius considers wheat produced in the U.S.
597
598 3.4. Harmonization

27
599 The procedure to harmonize parameters and assumptions generated similar impact values
600 calculated by the models (Figure 4). Note that all following numbers are presented in g
601 CO2eq per MJ of ethanol. Upon harmonization, default impacts for sugarcane ethanol (24.0
602 by GREET, 23.0 by BioGrace, and 43.9 by GHGenius) changed to values close to 16.1
603 calculated with VSB assumptions (17.5 by GREET, 17.3 by BioGrace, and 17.2 by
604 GHGenius) for ethanol produced in Brazil, representing a maximum variation of 8%. In terms
605 of corn ethanol, considering the average of dry and wet corn milling facilities, the results
606 obtained with default parameters (43.4 by BioGrace and 61.9 by GHGenius) changed to
607 values close to 57.7 calculated by GREET (57.0 by BioGrace and 56.2 by GHGenius),
608 achieving a maximum variation of 3%.
609 The results obtained in this work are in line with the previous harmonization study by Chum
610 and Warner [24]. In that study, impacts for corn ethanol produced in dry milling facilities
611 (43.4 by BioGrace and 62.8 by GHGenius) changed to 54.1 by BioGrace and 53.7 by
612 GHGenius within a maximum variation of results of 2.5% in comparison with the value of
613 52.5 by GREET.
28
614
615
616 Figure 4. Steps toward harmonization of public GHG emissions models for ethanol from
617 sugarcane (using VSB as a basis) in Brazil (top) and corn ethanol (using GREET as a basis)
618 in the U.S. (bottom).
619
620 3.5. Methodology comments
621 Although few LCA models and feedstocks were investigated to produce ethanol, the
622 conclusions from this study are broadly applicable to LCA work carried out in a
29
623 regulatory/policy context. The feedstocks selected for ethanol currently provide more than
624 90% of the ethanol produced and used globally [72].
625 Results of the harmonization emphasize that to understand the comparisons of LCA results
626 from commercial biofuels one must recognize: the geographic and temporal context for
627 feedstock agronomic production and of the performance of the industrial sector as
628 technologies used for biofuel production improved over time (as it is the case for the two
629 major feedstocks addressed here, as also shown by Chum et al. [73][74]; the product slate of
630 the biorefinery evolves; the approach for modeling land use and coproduct credits; broad
631 differences (e.g., uses natural gas instead of coal) or gaps (e.g., no pesticide or adoption of
632 different country agricultural production data for lack of specific country data) in inventories;
633 the scope and geographic context of the supply chain for chemicals, infrastructure, and
634 biofuel transport and use (can be different than the production).
635 Some models have been continually improved (e.g., GREET), including relevant commercial
636 pathways, while others lag, for instance, from 2013 GHGenius (used in the study) was
637 updated in 2018. Defaults data in BioGrace is out of date and may not reflect the current
638 commercial practices. However, BioGrace has a different purpose as it encourage users to
639 provide their own data.
640 Regarding sugarcane ethanol production in Brazil, for instance, calculation of the amount of
641 surplus electricity generated as a coproduct could be a variable. It is not an option in BioGrace
642 as the EU regulation does not consider this aspect. This is an example of the tool following
643 a directive that is not a research tool. This distinction could be clearer for researchers in the
644 model documentation.
645 Significant progress has been accomplished with the GREET.net improved graphic interface
646 that makes navigation easier for these complex models with intricately linked spreadsheets.
30
647 Continual updating is needed to decrease inconsistencies in comparing web and spreadsheet
648 results (e.g., field-burned straw). User-friendly interfaces for non-LCA experts would allow
649 users to choose among options (treatment of coproducts, transportation distances, type of
650 vehicles, etc.) to interpret results. Results obtained by the models presented by life-cycle
651 stages, for instance, would make visualization of the influence of specific inputs on results
652 more informative. This is just an example of presentation of results to facilitate the
653 comparison across models/conditions to identify most critical inputs to GHG impacts.
654
655 4. Conclusions
656 Our work shows that LCA tools calculate similar results when harmonized to the extent
657 possible. The results of the harmonization procedure appear encouraging in the context of
658 policies dealing with GHG emissions reduction targets, given that relatively few modification
659 steps of identified parameters and assumptions were used to generate similar GHG emissions
660 results for ethanol. On the other hand, the same fact shows that the models are highly sensitive
661 to these assumptions and methodological choices, which may be interpreted as a risk factor
662 in the policy arena.
663 The elaboration and application of the life-cycle GHG emissions assessment tools ultimately
664 align with the biofuels regulations or directives defined by governments individually or
665 regionally, which vary and may have different and/or conflicting requirements. Meanwhile,
666 different results obtained by using such models suggest that modeling tools should provide
667 transparent data sources and assumptions used in LCA calculations, to facilitate the
668 understanding of results in terms of geographical locations of production of biomass and
669 biofuels consumption. Regional directive requirements have harmonized data for specific
670 regions (EU); states have set specific tools for their use (California); or Canadian provinces
31
671 similarly have set up requirements as they implement and measure progress against stated
672 objectives. Requirements change as more scientific information emerges that justifies
673 changes. BioGrace is a public tool specifically designed for assessment of commercial
674 bioenergy systems in the European Union. The other three tools used started and continue to
675 be used as research tools, able to assess technology improvements (or feedstock supply
676 improvements), or to compare biofuel pathways in scenario analyses.
677 Review papers discussing biofuels GHG emission reductions must consider data sources and
678 clarify the models utilized, because their results depend on the purpose of the LCA, the
679 boundaries of the systems investigated, the stage of technology development, the source of
680 biomass production data (from small plots, field trial or commercial production), the
681 locations across the supply chain of biomass, and the use of resulting fuels.
682
683 5. Recommendations
684 This study recommends strengthened efforts to strive for transparency about the structures
685 and underlying assumptions of LCA models, their calculation mechanisms, and in the
686 reporting of results. To this end, around the world efforts should be redoubled within
687 stakeholder industries, the LCA/GHG modeler community, governmental entities, and multi-
688 lateral governmental and non-governmental organizations (e.g., CRC, Inc.) to work together
689 to increase transparency of models as well as to reach consensus on key areas such as
690 treatment of coproducts. Similarly, efforts to increase the understanding of climate effects on
691 emissions must continue to improve these tools (and many organizations are working towards
692 this end). This study mainly expresses the perspective of the LCA model user community,
693 not the view of a model developer.
32
694 It would facilitate comparisons if the structure of the models and their presentation followed
695 consolidated instructions for LCA studies (or well defined), providing clear definitions of
696 study scope and major functional unit metrics being assessed (e.g., GHG emissions as CO2eq
697 per MJ and/or per km) as well as explicit detail about the life-cycle dataset and inventories
698 used. This would also facilitate the understanding of the main assumptions of the models.
699 Similarly, built-in conversion of results from the individual model units to some common
700 functional unit to facilitate mutual understanding of results across models and regions, not
701 just by modelers but by the entire user and stakeholder communities.
702 Life-cycle inventories should be continually updated and improved to incorporate the most
703 data on biomass feedstock productivity, taking into account location-specific agricultural
704 practices as well as the vintage of industrial technologies being used. Research communities,
705 the other hand, should provide clear indication on the technology stage of development (and
706 thus level of confidence/uncertainty in data used), and biomass feedstock development stage
707 (for studies examining prospective future scenarios).
708
709 References
710
711 [1] De Jong S, Antonissen K, Hoefnagels R, Lonza L, Wang M, Faaij A, et al. Life-
712 cycle analysis of greenhouse gas emissions from renewable jet fuel production.
713 Biotechnol Biofuels 2017;10:1–18. doi:10.1186/s13068-017-0739-7.
714 [2] IEA. IEA Bioenergy Annual Report 2016. 2016 n.d.
715 https://www.ieabioenergy.com/publications/iea-bioenergy-annual-report-2016/
716 (accessed May 1, 2017).
717 [3] EPA US. Lifecycle Analysis of Greenhouse Gas Emissions under the Renewable
718 Fuel Standard 2016. https://www.epa.gov/renewable-fuel-standard-
719 program/lifecycle-analysis-greenhouse-gas-emissions-under-renewable-fuel
720 (accessed May 20, 2017).
721 [4] Hickox WH, Werner B, Gaffney P. Air Resources Board 2014:1–6.
722 [5] European Commission. Renewable Energy Directive 2017.
723 https://ec.europa.eu/energy/en/topics/renewable-energy/renewable-energy-directive
724 (accessed May 20, 2017).
725 [6] Macedo IC, Nassar AM, Cowiec AL, Seabra JEA, Marelli L, Otto M, et al.
33
726 Greenhouse Gas Emissions from Bioenergy. In: Souza, GM, Victoria, RL, Joly, CA,
727 Verdade L, editor. Bioenergy Sustain. Bridg. gaps, Sao Paulo: 2015, p. 583–616.
728 [7] (S&T)2 Consultants Inc. GHGenius Model 4.03 Volume 1: Model Background and
729 Structure 2013:445.
730 [8] (S&T)2 Consultants Inc. GHGenius Model 4.03 Volume 2: Data and Data Sources
731 2013:575.
732 [9] ANL. GREET Model 2017. https://greet.es.anl.gov/ (accessed April 20, 2017).
733 [10] EU. BioGrace 2015.
734 http://www.biograce.net/content/ghgcalculationtools/recognisedtool/ (accessed April
735 20, 2017).
736 [11] Canter CE, Dunn JB, Han J, Wang Z, Wang M. Policy Implications of Allocation
737 Methods in the Life Cycle Analysis of Integrated Corn and Corn Stover Ethanol
738 Production. Bioenergy Res 2016;9:77–87. doi:10.1007/s12155-015-9664-4.
739 [12] Cherubini F, Strømman AH, Ulgiati S. Influence of allocation methods on the
740 environmental performance of biorefinery products - A case study. Resour Conserv
741 Recycl 2011;55:1070–7. doi:10.1016/j.resconrec.2011.06.001.
742 [13] Wang L, Quiceno R, Price C, Malpas R, Woods J. Economic and GHG emissions
743 analyses for sugarcane ethanol in Brazil: Looking forward. Renew Sustain Energy
744 Rev 2014;40:571–82. doi:10.1016/j.rser.2014.07.212.
745 [14] Khatiwada D, Seabra J, Silveira S, Walter A. Accounting greenhouse gas emissions
746 in the lifecycle of Brazilian sugarcane bioethanol: Methodological references in
747 European and American regulations. Energy Policy 2012;47:384–97.
748 doi:10.1016/j.enpol.2012.05.005.
749 [15] Luo L, van der Voet E, Huppes G. Life cycle assessment and life cycle costing of
750 bioethanol from sugarcane in Brazil. Renew Sustain Energy Rev 2009;13:1613–9.
751 doi:10.1016/j.rser.2008.09.024.
752 [16] Koponen K, Soimakallio S, Kline KL, Cowie A, Brandão M. Quantifying the
753 climate effects of bioenergy – Choice of reference system. Renew Sustain Energy
754 Rev 2018;81:2271–80. doi:10.1016/j.rser.2017.05.292.
755 [17] Hill J, Polasky S, Nelson E, Tilman D, Huo H, Ludwig L, et al. Climate change and
756 health costs of air emissions from biofuels and gasoline. Proc Natl Acad Sci U S A
757 2009;106:2077–82. doi:10.1073/pnas.0812835106.
758 [18] Flugge MJ, Lewandrowski J, Rosenfeld J, Boland C, Hendrickson T, Jaglo K, et al.
759 A Life-Cycle Analysis of the Greenhouse Gas Emissions of Corn-Based Ethanol.
760 Rep Prep by ICF under USDA Contract No AG-3142-D-16-0243 2017:185.
761 [19] Wang M, Han J, Dunn JB, Cai H, Elgowainy A. Well-to-wheels energy use and
762 greenhouse gas emissions of ethanol from corn, sugarcane and cellulosic biomass for
763 US use. Environ Res Lett 2012;7. doi:10.1088/1748-9326/7/4/045905.
764 [20] Wang MQ, Han J, Haq Z, Tyner WE, Wu M, Elgowainy A. Energy and greenhouse
765 gas emission effects of corn and cellulosic ethanol with technology improvements
766 and land use changes. Biomass and Bioenergy 2011;35:1885–96.
767 doi:10.1016/j.biombioe.2011.01.028.
768 [21] Cavalett O, Chagas MF, Seabra JEA, Bonomi A. Comparative LCA of ethanol
769 versus gasoline in Brazil using different LCIA methods. Int J Life Cycle Assess
770 2013;18:647–58. doi:10.1007/s11367-012-0465-0.
771 [22] Seabra JEA, Macedo IC, Chum HL, Faroni CE, Sarto CA. Life cycle assessment of
772 Brazilian sugarcane products: GHG emissions and energy use. Biofuels, Bioprod
34
773 Biorefining 2011;5:519–32. doi:10.1002/bbb.289.
774 [23] Hennecke AM, Faist M, Reinhardt J, Junquera V, Neeft J, Fehrenbach H. Biofuel
775 greenhouse gas calculations under the European Renewable Energy Directive – A
776 comparison of the BioGrace tool vs. the tool of the Roundtable on Sustainable
777 Biofuels. Appl Energy 2013;102:55–62. doi:10.1016/j.apenergy.2012.04.020.
778 [24] Chum H, Warner E. Tools for greenhouse gas (GHG) assessment for biofuels: a
779 comparison. IEA Bioenergy Conf 2015.
780 http://ieabioenergy2015.org/fileadmin/veranstaltungen/2015/IEA_Bioenergy_Confer
781 ence/S08-4_Chum.pdf (accessed May 20, 2017).
782 [25] (S&T)2 Consultants I, Coordinating Research Council I. Transportation Fuel Life
783 Cycle Assessment: Validation and Uncertainty of Well-to-Wheel GHG Estimates
784 2013:293p.
785 [26] Unnasch S, Riffel B, Sanchez S, Waterland L. Review of Transportation Fuel Life
786 Cycle Analysis. 2011.
787 [27] Bonomi A, Cavalett O, Pereira da Cunha M, Lima MAP. Virtual Biorefinery. Cham:
788 Springer International Publishing; 2016. doi:10.1007/978-3-319-26045-7.
789 [28] International Organization for Standardization. ISO 14040. Environmental
790 management – Life-cycle assessment – Principles and framework 2006.
791 https://www.iso.org/standard/37456.html.
792 [29] Standardization IO for. ISO 14044. Environmental management – Life-cycle
793 assessment – Requirements and guidelines 2006.
794 https://www.iso.org/standard/38498.html.
795 [30] Delucchi M. A Lifecycle Emissions Model (LEM): Lifecycle emissions from
796 Transportation Fuels, Motor Vehicles, Transportation Modes, and Electricity 2003.
797 https://escholarship.org/uc/item/9vr8s1bb (accessed May 15, 2017).
798 [31] Joint European Commission. Well-to-wheels Analysis of Future Automotive Fuels
799 and Powertrains in the European Context 2014. http://iet.jrc.ec.europa.eu/about-jec/
800 (accessed April 15, 2017).
801 [32] Cavalett O, Junqueira TL, Dias MOS, Jesus CDF, Mantelatto PE, Cunha MP, et al.
802 Environmental and economic assessment of sugarcane first generation biorefineries
803 in Brazil. Clean Technol Environ Policy 2012;14:399–410. doi:10.1007/s10098-011-
804 0424-7.
805 [33] Pereira LG, Chagas MF, Dias MOS, Cavalett O, Bonomi A. Life cycle assessment of
806 butanol production in sugarcane biorefineries in Brazil. J Clean Prod 2015;96:557–
807 68. doi:10.1016/j.jclepro.2014.01.059.
808 [34] Pereira LG, Dias MOS, Junqueira TL, Chagas MF, Cavalett O, Filho RM, et al.
809 Production of butanol and other high valued chemicals using ethanol as feedstock
810 integrated to a first and second generation sugarcane distillery. Chem Eng Trans
811 2014;37:805–10. doi:10.3303/CET1437135.
812 [35] Pereira LG, Dias MOS, Junqueira TL, Pavanello LG, Chagas MF, Cavalett O, et al.
813 Butanol production in a sugarcane biorefinery using ethanol as feedstock. Part II:
814 Integration to a second generation sugarcane distillery. Chem Eng Res Des
815 2014;92:1452–62. doi:10.1016/j.cherd.2014.04.032.
816 [36] Pereira LG, Dias MOS, MacLean HL, Bonomi A. Investigation of uncertainties
817 associated with the production of n-butanol through ethanol catalysis in sugarcane
818 biorefineries. Bioresour Technol 2015;190:242–50.
819 doi:10.1016/j.biortech.2015.04.095.
35
820 [37] Pereira LG, Dias MOS, Mariano AP, Maciel Filho R, Bonomi A. Economic and
821 environmental assessment of n-butanol production in an integrated first and second
822 generation sugarcane biorefinery: Fermentative versus catalytic routes. Appl Energy
823 2015;160:120–31. doi:10.1016/j.apenergy.2015.09.063.
824 [38] Moraes BS, Junqueira TL, Pavanello LG, Cavalett O, Mantelatto PE, Bonomi A, et
825 al. Anaerobic digestion of vinasse from sugarcane biorefineries in Brazil from
826 energy, environmental, and economic perspectives: Profit or expense? Appl Energy
827 2014;113:825–35. doi:10.1016/j.apenergy.2013.07.018.
828 [39] Cardoso T de F, Cavalett O, Chagas MF, Morais ER de, Carvalho JLN, Franco HCJ,
829 et al. Technical and economic assessment of trash recovery in the sugarcane
830 bioenergy production system. Sci Agric 2013;70:353–60. doi:10.1590/S0103-
831 90162013000500010.
832 [40] Cavalett O, Chagas MF, Junqueira TL, Watanabe MDB, Bonomi A. Environmental
833 impacts of technology learning curve for cellulosic ethanol in Brazil. Ind Crops Prod
834 2017;106:31–9. doi:10.1016/j.indcrop.2016.11.025.
835 [41] Junqueira TL, Chagas MF, Gouveia VLR, Rezende MCAF, Watanabe MDB, Jesus
836 CDF, et al. Techno-economic analysis and climate change impacts of sugarcane
837 biorefineries considering different time horizons. Biotechnol Biofuels 2017;10:50.
838 doi:10.1186/s13068-017-0722-3.
839 [42] Souza A, Watanabe MDB, Cavalett O, Ugaya CML, Bonomi A. Social life cycle
840 assessment of first and second-generation ethanol production technologies in Brazil.
841 Int J Life Cycle Assess 2018;23:617–28. doi:10.1007/s11367-016-1112-y.
842 [43] Watanabe MDB, Chagas MF, Cavalett O, Guilhoto JJM, Griffin WM, Cunha MP, et
843 al. Hybrid Input-Output Life Cycle Assessment of First- and Second-Generation
844 Ethanol Production Technologies in Brazil. J Ind Ecol 2016;20:764–74.
845 doi:10.1111/jiec.12325.
846 [44] Dias MOS, Junqueira TL, Cavalett O, Cunha MP, Jesus CDF, Rossell CEV, et al.
847 Integrated versus stand-alone second generation ethanol production from sugarcane
848 bagasse and trash. Bioresour Technol 2012;103:152–61.
849 doi:10.1016/j.biortech.2011.09.120.
850 [45] Mariano AP, Dias MOS, Junqueira TL, Cunha MP, Bonomi A, Filho RM. Butanol
851 production in a first-generation Brazilian sugarcane biorefinery: Technical aspects
852 and economics of greenfield projects. Bioresour Technol 2013;135:316–23.
853 doi:10.1016/j.biortech.2012.09.109.
854 [46] Mariano AP, Dias MOS, Junqueira TL, Cunha MP, Bonomi A, Filho RM.
855 Utilization of pentoses from sugarcane biomass: Techno-economics of biogas vs.
856 butanol production. Bioresour Technol 2013;142:390–9.
857 doi:10.1016/j.biortech.2013.05.052.
858 [47] Dias MOS, Pereira LG, Junqueira TL, Pavanello LG, Chagas MF, Cavalett O, et al.
859 Butanol production in a sugarcane biorefinery using ethanol as feedstock. Part I:
860 Integration to a first generation sugarcane distillery. Chem Eng Res Des
861 2014;92:1441–51. doi:10.1016/j.cherd.2014.04.030.
862 [48] IPCC. Climate Change 2001: The Scientific Basis. Cambridge University Press;
863 2001.
864 [49] IPCC. Climate Change 2007: The Physical Science Basis. New York: Cambridge
865 University Press; 2007.
866 [50] IPCC. Climate Change 2013: The Physical Science Basis. New York: Cambridge
36
867 University Press; 2013.
868 [51] IPCC. The IPCC Scientific Assessment. New York: Cambridge University Press;
869 1990.
870 [52] IPCC. Climate Change 1995: The Science of Climate Change. New York:
871 Cambridge University Press; 1995.
872 [53] ANL. Carbon Calculator for Land Use Change from Biofuels Production (CCLUB)
873 2015. https://greet.es.anl.gov/publication-cclub-manual (accessed May 5, 2017).
874 [54] Taheripour F, Tyner WE WM. Global land use change due to the U.S. cellulosic
875 biofuels program simulated with the GTAP model 2011.
876 https://greet.es.anl.gov/files/luc_ethanol (accessed May 5, 2017).
877 [55] van Deusen PC HL. COLE web applications suite 2013.
878 http://www.ncasi2.org/COLE/ (accessed May 3, 2017).
879 [56] Harris NL, Grimland S BS. Land use change and emission factors: Updates since
880 proposed RFS rule. 2009.
881 [57] Kwon H-Y, Hudson RJM. Quantifying management-driven changes in organic
882 matter turnover in an agricultural soil: An inverse modeling approach using
883 historical data and a surrogate CENTURY-type model. Soil Biol Biochem
884 2010;42:2241–53. doi:10.1016/j.soilbio.2010.08.025.
885 [58] Finnveden G, Hauschild MZ, Ekvall T, Guinée J, Heijungs R, Hellweg S, et al.
886 Recent developments in Life Cycle Assessment. J Environ Manage 2009;91:1–21.
887 doi:10.1016/j.jenvman.2009.06.018.
888 [59] Ekvall T, Weidema BP. System boundaries and input data in consequential life cycle
889 inventory analysis. Int J Life Cycle Assess 2004;9:161–71.
890 doi:10.1007/BF02994190.
891 [60] Guinée JB, Gorrée M, Heijungs R, Huppes G, Kleijn R KA de. Handbook on life-
892 cycle assessment. Operational guide to the ISO standards. Dordrecht: Kluwer
893 Academic Publishers; 2002.
894 [61] Heijungs R, Guinée JB. Allocation and ‘what-if’ scenarios in life cycle assessment of
895 waste management systems. Waste Manag 2007;27:997–1005.
896 doi:10.1016/j.wasman.2007.02.013.
897 [62] Wang M, Huo H, Arora S. Methods of dealing with co-products of biofuels in life-
898 cycle analysis and consequent results within the U.S. context. Energy Policy
899 2011;39:5726–36. doi:10.1016/j.enpol.2010.03.052.
900 [63] International Organization for Standardization. ISO 13065. Sustainability Criteria for
901 Bioenergy 2015. https://www.iso.org/standard/52528.html (accessed May 24, 2017).
902 [64] Ecoinvent. Ecoinvent v2.2 2010. http://www.ecoinvent.org/database/older-
903 versions/ecoinvent-version-2/reports-on-ecoinvent-2/reports-on-ecoinvent-2.html
904 (accessed January 23, 2017).
905 [65] Stehfest E, Bouwman L. N2O and NO emission from agricultural fields and soils
906 under natural vegetation: summarizing available measurement data and modeling of
907 global annual emissions. Nutr Cycl Agroecosystems 2006;74:207–28.
908 doi:10.1007/s10705-006-9000-7.
909 [66] IPCC. N2O Emissions From Managed Soils, and CO2 Emissions From Lime and
910 Urea application. Guidel Natl Greenh Gas Invent 2006:54.
911 [67] Macedo, IC, Leal MRLV, da Silva J. Greenhouse gases emissions in the production
912 and use of ethanol from sugarcane in Brazil (in Portuguese) 2004.
913 www.unica.com.br/download.php?idSecao=17&id=11168105 (accessed September
37
914 17, 2017).
915 [68] Macedo IC, Seabra JEA, Silva JEAR. Green house gases emissions in the production
916 and use of ethanol from sugarcane in Brazil: The 2005/2006 averages and a
917 prediction for 2020. Biomass and Bioenergy 2008;32:582–95.
918 doi:10.1016/j.biombioe.2007.12.006.
919 [69] UDOP. Legislation for sugarcane straw burning (in Portuguese) n.d.
920 http://www.udop.com.br/index.php?item=legislacao1&op=40 (accessed April 1,
921 2017).
922 [70] Wang M, Wu M, Huo H. Life-cycle energy and greenhouse gas emission impacts of
923 different corn ethanol plant types. Environ Res Lett 2007;2:024001.
924 doi:10.1088/1748-9326/2/2/024001.
925 [71] U.S. Department of Agriculture (USDA). Economic Research Service (ERS): Corn
926 and other feedgrains 2011. https://www.ers.usda.gov/topics/crops/corn/ (accessed
927 December 12, 2017).
928 [72] U.S. Department of Energy (USDoE). Maps and Data - Global Ethanol Production
929 2016. https://www.afdc.energy.gov/data/10331 (accessed September 17, 2018).
930 [73] Chum HL, Zhang Y, Hill J, Tiffany DG, Morey RV, Goss Eng A, et al.
931 Understanding the evolution of environmental and energy performance of the US
932 corn ethanol industry: evaluation of selected metrics. Biofuels, Bioprod Biorefining
933 2014;8:224–40. doi:10.1002/bbb.1449.
934 [74] Chum HL, Warner E, Seabra JEA, Macedo IC. A comparison of commercial ethanol
935 production systems from Brazilian sugarcane and US corn. Biofuels, Bioprod
936 Biorefining 2014;8:205–23. doi:10.1002/bbb.1448.
937
938 Declarations
939
940 Competing interests
941 The authors declare no competing financial interests.
942
943 Funding
944 Financial support for this work was provided by Task 39 Commercializing Liquid Fuels of
945 the International Energy Agency (IEA) and by the U.S. Department of Energy Bioenergy
946 Technologies Office to the Alliance for Sustainable Energy, manager of NREL under
947 contract No. DE-AC36-08GO28308 and the Bioenergy Technologies Office projects
948 International Sustainability and IEA Bioenergy Task 38.
949
950 Authors' contributions
951 All six authors contributed equally to the conception and design of this study, interpretation
952 of the results, and editing of the text.
953
954 Acknowledgements
955 This study was developed as a collaboration between the Brazilian Bioethanol Science and
956 Technology Laboratory (CTBE) and the National Renewable Energy Laboratory (NREL) for
957 phase 1 of the project entitled “Comparison of Biofuel Life-cycle Assessment Tools”
958 prepared for Task 39 of the International Energy Agency (IEA) Bioenergy Technology
959 Collaboration Program (IEA Bioenergy TCP). The authors express their gratitude to Jim
960 McMillan from NREL, Jack Saddler, Susan van Dyk, and, Anna Ringsred from the
38
961 University of British Columbia (UBC), John Neeft from the Rijksdienst voor Ondernemend
962 (RVO), Horst Feherenbach (Institute for Energy and Environmental Research (IFEU)), Don
963 O’Connor from (S&T)2 Consultants, and Jennifer Dunn, Michael Wang, Jeongwoo Han, and
964 Hao Cai from Argonne National Laboratory (ANL). Also acknowledged are Patrick Lamers
965 and Kathy Cisar from NREL. The authors are grateful to acknowledge helpful guidance
966 provided by Kristen Johnson and Jim Spaeth from the U.S. Department of Energy’s
967 Bioenergy Technologies Office (BETO) and IEA Bioenergy Task 38’s Annette Cowie (NSW
968 Department of Primary Industries) and Miguel Brandão (IEA Bioenergy Task 38).
969
970 NOTE:
971 This work was authored in part by the National Renewable Energy Laboratory, operated by
972 Alliance for Sustainable Energy, LLC, for the U.S. Department of Energy (DOE) under
973 Contract No. DE-AC36-08GO28308. Funding provided by the Bioenergy Technologies,
974 U.S. Department of Energy Office of the Energy Efficiency and Renewable Energy Solar
975 Energy Technologies Office. The views expressed in the article do not necessarily represent
976 the views of the DOE or the U.S. Government. The U.S. Government retains and the
977 publisher, by accepting the article for publication, acknowledges that the U.S. Government
978 retains a nonexclusive, paid-up, irrevocable, worldwide license to publish or reproduce the
979 published form of this work, or allow others to do so, for U.S. Government purposes.
39

Comparison of Biofuel Life-Cycle GHG Emissions Assessment Tools: The Case Studies of Ethanol Produced From Sugarcane, Corn, and Wheat

Uploaded by

Copyright:

Available Formats

Comparison of Biofuel Life-Cycle GHG Emissions Assessment Tools: The Case Studies of Ethanol Produced From Sugarcane, Corn, and Wheat

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Comparison of Biofuel Life-Cycle GHG Emissions Assessment Tools: The Case Studies of Ethanol Produced From Sugarcane, Corn, and Wheat

Uploaded by

Copyright:

Available Formats

1 Comparison of biofuel life-cycle GHG emissions assessment tools: the case

2 studies of ethanol produced from sugarcane, corn, and wheat

7 Materials, Campinas, SP, Brazil

99 reducing GHG emissions of increased importance [1–6].

105 pathways for fuels coupled to vehicle life-cycle systems.

115 (FQD), with harmonized data for EU nations.

161 automotive equipment.

173 2.1. Life cycle assessment

189 assumptions utilized.

195 2.2. Considered LCA models

198 presented in section 3.

200 2.2.1. GREET

211 2016 was used in this study.

214 purposes and updated periodically.

216 2.2.2. GHGenius

225 this study was completed.

227 2.2.3. BioGrace

233 version was used).

242 with the EU-RED.

244 2.2.4. Virtual Sugarcane Biorefinery

254 wages, educational level, and gender distribution) aspects [27,32–47].

265 year (2011).

302 Brazil and of GREET for corn ethanol in the U.S.

304 3. Results and discussion

305 3.1. Comparison of the models

309 N2O emissions).

343 2007 report [49].

358 decisions [59].

382 into the consequential LCA approach.

388 the consequences of using alternative procedures.

396 substitution or other allocation methods.

398 3.1.2. Upstream life-cycle data

404 [31] and the Ecoinvent database v2.2 [64], respectively.

413 the VSB.

421 residues application rates.

424 of N as N2O, at 1% of synthetic N fertilizer application and 2% for manure; volatilization of

430 of NH3 and nitrate to N2O).

439 sugarcane and 0.9% for corn).

442 3.2. Life-cycle inventories

447 of the supplementary material.

450 3.2.1. Sugarcane ethanol production in Brazil

464 publication projected rates to 2020.

477 be manually set by the user.

481 Brazil to the destination country.

483 3.2.2. Corn ethanol

496 2011 [71].

500 corn and ethanol production and transportation.

502 3.2.3. Wheat ethanol

516 3.3. Comparison of calculated GHG emissions

542 GREET calculates them.

553 (Table 3).

560 total impact in GREET (energy allocation), 0% in BioGrace (energy allocation), 9% in

576 sugarcane harvest) [39].

598 3.4. Harmonization

608 achieving a maximum variation of 3%.

613 52.5 by GREET.

620 3.5. Methodology comments

624 90% of the ethanol produced and used globally [72].

639 provide their own data.