Scholarly article on topic 'Detailed life cycle assessment of Bounty® paper towel operations in the United States'

Detailed life cycle assessment of Bounty® paper towel operations in the United States Academic research paper on "Chemical sciences"

CC BY-NC-ND
0
0
Share paper
Academic journal
Journal of Cleaner Production
OECD Field of science
Keywords
{"Consumer product" / "Paper towel" / "Sustainability metrics" / "Sustainability indicators" / "Life cycle assessment" / "Life cycle impact assessment"}

Abstract of research paper on Chemical sciences, author of scientific article — Wesley Ingwersen, Maria Gausman, Annie Weisbrod, Debalina Sengupta, Seung-Jin Lee, et al.

Abstract Life Cycle Assessment (LCA) is a well-established and informative method of understanding the environmental impacts of consumer products across the entire value chain. However, companies committed to sustainability are interested in more methods that examine their products and activities' impacts. Methods that build on LCA strengths and illuminate other connected but less understood facets, related to social and economic impacts, would provide greater value to decision-makers. This study is a LCA that calculates the potential impacts associated with Bounty® paper towels from two facilities with different production lines, an older one (Albany, Georgia) representing established technology and the other (Box Elder, Utah), a newer state-of-the-art platform. This is unique in that it includes use of Industrial Process Systems Assessment (IPSA), new electricity and pulp data, modeled in open source software, and is the basis for the development of new integrated sustainability metrics (published separately). The new metrics can guide supply chain and manufacturing enhancements, and product design related to environmental protection and resource sustainability. Results of the LCA indicate Box Elder had improvements on environmental impact scores related to air emission indicators, except for particulate matter. Albany had lower water use impacts. After normalization of the results, fossil fuel depletion is the most critical environmental indicator. Pulp production, electricity, and fuels for product production drive fossil fuel depletion. Climate change, land occupation, and particulate matter are also relevant. Greenhouse gas (GHG) emissions by pulp, electricity, papermaking, and landfill methane from the disposed product, drive climate change impacts. Pulp provides significant offsets to balance climate change impacts due to sequestration of atmospheric carbon dioxide. Ninety-nine percent of land occupation is for the growth of the trees for pulp production. Papermaking, electricity, and pulp production cause the most potential particular matter formation.

Academic research paper on topic "Detailed life cycle assessment of Bounty® paper towel operations in the United States"

Accepted Manuscript

Detailed Life Cycle Assessment of Bounty Paper Towel Operations in the United States

Wesley Ingwersen, Maria Gausman, Annie Weisbrod, Debalina Sengupta, Seung-Jin Lee, Jane Bare, Ed Zanoli, Gurbakash S. Bhander, Manuel Ceja

PII: S0959-6526(16)30438-3

DOI: 10.1016/j.jclepro.2016.04.149

Reference: JCLP 7174

To appear in: Journal of Cleaner Production

Received Date: 5 November 2015

Revised Date: 28 April 2016

Accepted Date: 29 April 2016

Please cite this article as: Ingwersen W, Gausman M, Weisbrod A, Sengupta D, Lee S-J, Bare J, Zanoli

E, Bhander GS, Ceja M, Detailed Life Cycle Assessment of Bounty Paper Towel Operations in the United States, Journal of Cleaner Production (2016), doi: 10.1016/j.jclepro.2016.04.149.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Detailed Life Cycle Assessment of Bounty® Paper Towel Operations in the United States

1 2* 2 13

Wesley Ingwersen , Maria Gausman , Annie Weisbrod , Debalina Sengupta , , Seung-Jin Lee 1 4, Jane Bare 1, Ed Zanoli 2, Gurbakash S. Bhander 1, and Manuel Ceja2

1 U.S. Environmental Protection Agency, National Risk Management Research Laboratory, Cincinnati, OH 45238

The Procter & Gamble Company, Global Product Stewardship and Product Supply, Cincinnati, OH 45224

Texas A&M University, Artie McFerrin Department of Chemical Engineering, College Station, TX 77843

4 University of Michigan, Department of Earth and Resource Science, Flint, MI 48502. * Corresponding Author E-Mail: gausman.mm@pg.com Received: / Accepted: / Published:

Abstract

Life Cycle Assessment (LCA) is a well-established and informative method of understanding the environmental impacts of consumer products across the entire value chain. However, companies committed to sustainability are interested in more methods that examine their products and activities' impacts. Methods that build on LCA strengths and illuminate other connected but less understood facets, related to social and economic impacts, would provide greater value to decision-makers. This study is a LCA that calculates the potential impacts associated with Bounty® paper towels from two facilities with different production lines, an older one (Albany, Georgia) representing established technology and the other (Box Elder, Utah), a newer state-of-the-art platform. This is unique in that it includes use of Industrial Production Systems Assessment (IPSA), new electricity and pulp data, modeled in open source software, and is the basis for the development of new integrated sustainability metrics (published separately). The new metrics can guide supply chain and manufacturing enhancements, and product design related to environmental protection and resource sustainability. Results of the LCA indicate Box Elder had improvements on environmental impact scores related to air emission indicators, except for particulate matter. Albany had lower water use impacts. After normalization of the results, fossil fuel depletion is the most critical environmental indicator. Pulp production, electricity, and fuels for product production drive fossil fuel depletion. Climate change, land occupation, and particulate matter are also relevant. Greenhouse gas (GHG) emissions by pulp, electricity, papermaking, and landfill methane from the disposed product, drive climate change impacts. Pulp provides significant offsets to balance climate change impacts due to sequestration of atmospheric carbon dioxide. Ninety-nine percent of land occupation

is for me growth oi nie uees for puip production. rapeimaking, eiecuicuy, anu puip production cause the most potential particular matter formation.

Keywords: consumer product; paper towel; sustainability metrics; sustainability indicators; life cycle assessment; life cycle impact assessment

1. Introduction

Procter & Gamble is a multi-billion dollar consumer products company that incorporates sustainability in its purpose (Procter & Gamble, 2015). P&G's sustainability vision includes powering its plants with 100% renewable energy, using 100% renewable and/or recycled materials in products and packaging, having 0% consumer and manufacturing waste go to landfill, and designing delightful consumer products while maximizing the conservation of resources. The company employs many tools, innovations, and experts to make progress toward this vision.

For example, in January 2012, the US EPA's Office of Research and Development's National Risk Management Research Laboratory (NRMRL) and P&G signed a 5-year Cooperative Research and Development Agreement (CRADA) to support the development of methods and tools for sustainability assessment within consumer product life cycles. An initial result has been a sustainability assessment approach that incorporates LCA with the use of novel integrated metrics (Ingwersen et al., 2014; Young et al., 2012). This work builds on two decades of LCA experience with US EPA and P&G scientists in research, method development, application, and collaboration through the professional society, "Society of Environmental Toxicology and Chemistry (SETAC)" and local projects (Curran, 2000; Fava et al., 1991; Jolliet et al., 2014b; Jolliet et al., 2004b; Margni et al., 2007); while providing EPA's researchers the opportunity to develop and apply tools, including newly developed LCA software (Gooch and Mack, 2012).

This collaborative effort addresses the following questions:

1) What product assessment approaches provide accurate and actionable information about social, economic, and environmental pillars of sustainability?

2) Can these approaches inform how changes in product design and manufacturing influence these pillars up and down the supply chain?

3) Is there an assessment framework that puts it all together practically, efficiently, and is actionable in any function and for any product?

P&G selected paper family products, and in particular paper towels, as the subject of an initial targeted effort to address these questions.

This paper towel LCA study is a key objective of the effort, where its results serve as the basis for the integrated sustainability metrics (Ingwersen et al., 2016). P&G has a history of using LCA to assess products and guide innovation (Saouter et al., 2002; McDougall et al., 2008; Weisbrod and Loftus, 2012; Weisbrod and Van Hoof, 2012; Van Hoof et al., 2014). The above questions stem from that experience.

llllO 10 U-l.

sustainability science with real application to consumer products with global supply chains (Ingwersen et al., 2014; Weisbrod et al., 2013; Young et al., 2012). One example of the CRADA results, incorporated and further tested in this LCA, is the Industrial Process System Assessment (IPSA) methodology (Sengupta et al., 2015a). IPSA is a multiple step allocation approach for connecting information from the production line level up to the facility level, and vice versa, using a multiscale model of process systems (Sengupta et al., 2015a). The method helps resolve challenges in assessing multi-product or multi-production line systems (Bousquin et al., 2010).

While this LCA model is foundational to exploring new integrated sustainability metrics, it is also designed to stand-alone. It thoroughly evaluates different supply chains and production processes used by two lines making paper towels at two facilities in the U.S. in 2012. The lines differ by age, location (Albany, Georgia vs. Box Elder, Utah), and technology. Other studies published on paper products do not have the resolution of assessing product impacts to the level of a single production line in a multiline/product manufacturing facility (Madsen, 2007; Montalbo et al, 2011; Joseph et al., 2015; Boguski, T.K., 2010). It incorporates detailed data from the IPSA allocation method, updated data on regional US electricity grids and pulp production, and a normalization procedure to identify key processes and materials that contribute most to potential environmental impacts in the life cycle of a consumer good. P&G benefits from the LCA results for having the most accurate and recent model of Bounty to use as a benchmark to compare innovations to. It also serves to compare the value of conducting a complex LCA relative to other more simple methods, such as tracking energy use and waste generation (i.e., would the same priorities to improve processes or materials be identified by a cost and resource analysis as by an LCA?).

For the average LCA practitioner, this study demonstrates a number of important advances that can improve the quality of LCA studies. These include the benefits of using the IPSA method in comparison with conventional facility-level allocation approaches, the benefits of using newly-developed US regionalized electricity life cycle inventory, the use of an openLCA version of the EPA Waste Reduction Model (WARM), and the use of new functionality for advanced scenario modeling in openLCA software, all of which are not described elsewhere. It also provides a cradle-to-grave paper towel life cycle inventory of high data quality from a major global manufacturer with detailed contribution analysis describing the results from life cycle impact assessment.

2. Materials and methods

The scope of this analysis is cradle-to-grave. Foreground data account for forestry and wood chip production, pulp production, papermaking and converting, multiple transportation steps, and end-of life of all wastes. The Life Cycle Inventory (LCI) data are aggregated at the following levels: forestry and pulp production, pulp transport, paper towel production, fuels for each paper towel facility, electricity for each facility, product distribution, and end-of-life. Product use by a consumer is assumed not to lead to any measureable impact.

The study is conducted on one roll of Bounty regular towels. The unit is not based on function, because there are many ways that people commonly use paper towels. The two lines being compared

lui ^uiiipamig piu^csscs ani

equipment function is more useful.

An iterative approach was taken where a preliminary screening-level LCA was performed in order to anticipate important sources of impact and direct subsequent work. The analysis was performed according to the same methods described here, using preliminary inventory data representing a single limited supply chain with primarily secondary data from the US Life Cycle Inventory (USLCI) (NREL, 2013) and Ecoinvent 2.2 (ecoinvent Centre, 2010). Preliminary results indicated the potential importance of pulp and direct energy consumption (particularly electricity) at the papermaking facility to many impact categories, and the importance of allocation choices/assumptions (Ingwersen et al., 2013). Therefore, additional efforts were made to improve the data quality of the pulp and electricity LCI, and to develop a more accurate method for allocating impacts from the papermaking facility to the Bounty product of interest.

The LCA was conducted by both EPA and P&G researchers working with EXCEL® spreadsheets and openLCA framework 1.4 (© 2007-2015 GreenDelta). ISO 14040 standards are followed and the highest level of data quality is utilized whenever available (ISO, 2006).

2.1. Inventory Modeling

Original LCI data from suppliers and P&G for pulp production and transport, paper towel production, and product distribution are used in the study.

Pulp production and transport

Forestry —Logs—►

Chipping Chips

Pulping Pulp

Papermaking

and Converting

Pulp Mill

Paper Facility

Figure 1. Generic processes are shown for pulp production, papermaking, and converting. Each box represents a separate unit process for the pulp LCIs used in this study.

Pulp production is represented by a set of unit processes for Kraft pulp (almost pure cellulose) used for paper towels, referred to henceforth as Pulp. Figure 1 shows the basic steps in pulp production from harvest through arrival at the paper facility. Transportation between the processes is not pictured, but is included in data for forestry, sawmill, pulp mill, and paper facilities.

The original pulp data were secured from multiple suppliers in several countries from 2010 to 2014. These confidential LCIs cover material sourcing and production beginning with forestry operations through pulp production. Not included in LCIs, but an important consideration for data reliability and appropriateness for environmental indices, is that independent third-party verification systems (e.g.,

sustainable forest management and wood traceability. P&G works with global multi-stakeholder organizations (e.g. World Wildlife Fund) on the development of tools and scientific methods to protect both the commercial value and services that forests provide, such as biodiversity, watershed protection, and climate moderation (Procter & Gamble, 2015).

Pulp mills commonly acquire a large fraction of their non-electricity energy needs from combustion of biomass fuels like wood chips and recovered fuels like black liquor that are of biogenic origin. Emissions of carbon dioxide from biogenic sources are typically excluded from facility reporting in systems such as EPA's GHG eGRET tool (US EPA, 2015). This study includes biogenic CO2 emissions in the inventory, as recommended by a recently developed product category rule for market pulp (FP Innovation, 2015). Therefore, CO2 emissions were determined by calculations based on standard methods using the carbon content of the reported fuels used (USEPA, 2015).

Forestry and sawmills are represented with data generated by the Consortium for Research on Renewable Industrial Materials (CORRIM) (Puettmann et al., 2010; Wagner et al., 2009), except for data on land occupation/conversion and carbon uptake. The CORRIM data include processes to grow, fell, delimb, skid, load timber onto a truck, and replant following harvest. The wood chips and forest residues for pulp production for the paper towels come from multiple sources and co-product generation processes, such as lumber and mill scrap and on-site chipping. Data from CORRIM do not separately identify the different sources of wood chips that become the pulp. A very conservative approach is used for this model and the trees are assumed to be grown exclusively for pulp production with no allocation for other uses. Under confidentiality disclosure agreements, land occupation and conversion data were collected and averaged from multiple growers as primary sources used by the pulp mills. Carbon uptake was calculated using standard IPCC methods based on forest species harvested, harvest density, wood density; assuming 0.5103 kg C/kg oven dry wood (IPCC 2007). Age of trees at harvest ranged from 6-68 years and harvest volume 40-219 dry m /hectare (Binkley, 2014; Cochran and Dahms, 2000; USFS, 1983). Wood densities ranged from 445-574 kg/m3 at 12% moisture. The burning of plantation residue following harvest is assumed to not occur, as consistent with the third party verification standards. In accordance with product carbon foot printing standards (BSI, 2011; EC, 2013; GHG Protocol, 2011), only the carbon sequestered in the trees that were harvested was accounted for in carbon uptake; other carbon uptake associated with forest land was excluded. For the portion of carbon uptake that was included, emissions of CO2 from combustion and decay of biogenic sources were included across all phases of the life cycle. With this method for developing inventory, all inputs and outputs to processes are tracked to achieve a carbon balance.

Wood residuals from debarking and chipping are assumed to be collected for use as fuel for pulp production, and allocation was performed on a mass-basis for these processes. For Albany, sawmills provide some waste residues as fuels, but not at Box Elder.

The pulp is made in mills via a thermo-chemical process from wood chips. Inputs to pulping include various fuel sources, purchased electricity, processing chemicals, and water. To represent the pulping process, data were solicited from 2010-2014 from seven pulp mills. Transportation distances from

1CII VC5L lUvClUVJII LU pUlJJ 111.

v W V ^OUiiiClL^'U UClij^U Wll UCllCl 11 Will 111V,/ V/U111111UU1

k_J U1 V ^

2009), and mode was assumed to be heavy-duty tractor trailer.

Paper towel production

Paper towel production includes both papermaking and paper converting, including embossing, rolling paper on cores and putting rolls in primary packaging. The papermaking and converting lines operate independently. Towels are tracked by a specific pair of making and converting lines (a line pair). Two line pairs, each from a different North American facility, were thus chosen based on their operation for producing the same paper substrate, i.e. Bounty regular, over the same time period, differences in papermaking and converting technologies, and differences in facility characteristics. Table 1 summarizes some of the key differences between the selected lines.

Table 1. Characteristics of Bounty® Lines Selected for Case Study

Aspect Albany Facility/Line A Box Elder Facility/Line B

Line Technology Older traditional platform Newest state-of-the-art platform

Primary Fuel & Energy Sources Natural gas, Biomass, Grid Electricity Natural gas, Grid Electricity

Comparative Facility Size Large Small

Facility Age > 30 years old < 10 years old

Primary Emission Control Technology BACT Combustion, Separators & Scrubbers, Wet ESP, Bag & Drum Filtration Low-NOx Combustion, Separators & Scrubbers, Drum Filtration

BACT = Best Available Control Technology; ESP = Electrostatic Precipitator; NOx = nitrogen oxides

Conventional facility level product allocation by mass or value did not provide the needed accuracy to obtain accurate inventories representative of the line pairs. Totally avoiding allocation was impossible because of confidentiality, lack of line level monitoring of some inputs and outputs, and because of the need to attribute some of the processes serving multiple lines (e.g. utilities) to a roll from a specific line pair. IPSA is a structured method for assessing the inputs and outputs related to a product of interest when the intention is to compare products from specific production lines in one or more complex industrial facilities (Figure 2). It uses sub-process modeling to avoid allocation when data are available at a sub-process level, and provide clear allocation when not available, within the context of a structured approach. In Step 1, a full list of flows into and out of the facility including material, energy, products, and releases were obtained from the facilities. Information was obtained on equipment and throughput to model capacities. In Step 2, flows were assigned to direct process (papermaking and converting); ancillary (e.g., boilers) and non-process (e.g., storage space) based on equipment usage data. In Step 3, flows further split among the direct processes to papermaking and converting line pairs, and among the ancillary processes based on equipment type. In Step 4, flows are assigned to specific papermaking or converting lines. Finally, flow amounts assigned to each component of the system including those for the line pair, the line pairs use of ancillary equipment, and

the non-process activities aie

produced by the line pair.

icu ana uiViucu acioss tue total mass 01

UUUCl Ul Híleles I

Step 1: Facility Flows

Materials

Utilities

Facility

Products

Releases

_Step 2: Process Classification

Materials

Direct Process

Utilities

Ancillary Process

Non-Process

Products

Releases

Step 3: Production System Definition

Materials

UIIKiCS

Direct Process

Prod<jzt\on ¿retenu Production Syïtom 1 Production System 2

Ancillary Process

tooo-Process

Products

Releases

Materials

Utilities

Step 4 Allocation at Multiple Scales

Direct Process Production Systems

Production System l I Linel

Line 2

Linen»

Line M

Production System 2 line 1 I

Une 2 I

line W

Production System X Linel

line 2

Ancillary Process

Non-Process

Products

Releases

Figure 2. IPSA four-step analysis method from (Sengupta et al., 2015a).

Fuels and electricity for paper towel facilities

For modeling electricity production for the US-based facilities for pulp and paper towel production, new LCI was developed for US eGRID (multi-state) regions (Lee et al., 2015). There is relatively high-energy demand for pulp and paper production (EPA, 2010), and electricity production is a potentially significant source of impacts (Boguski, 2010; Ingwersen et al., 2012) in this sector. Known regional differences in electricity-related impacts are influential (Mutel et al., 2011). For pulp production overseas, comparable inventories based on national mixes were developed. When the region for a supply chain was unknown, the US average electricity production model was used.

For developing fuel-specific electricity processes (e.g., electricity from coal), electricity unit processes were aggregated by fuel source from Ecoinvent 2.2 data, maintained their inputs, and replaced the emissions associated with combustion for electricity production with fuel source-specific emission factors based on US electricity. Fuel-source specific emissions factors for three GHGs (CO2, CH4, N2O) and the criteria air pollutants (CO, NOx, PM2.5, PM10, SOx, VOCs) were adopted from Cai et al. (2012), which are the same used in the GREET model. Fuel source-specific water loss estimates were included based on Macknick (Macknick et al., 2012). Electricity processes by fuel source were then included as inputs into regional and/or national level electricity generation mix processes. Regional US

US, national level mixes came from international energy statistics available from the Energy Information Administration (EIA, 2015). Table A1 in the Appendix summarizes the power mixes assumed for the facilities.

Electricity generation mix processes were then connected to 'electricity, at industrial user' processes for modeling distribution to point of use. National level losses associated with distribution were estimated based on Schmidt et al. (2015). No emissions or infrastructure demands were modeled for distribution, assuming the insignificance of impacts associated with distribution (aside from losses) relative to production and other upstream processes.

Packaging

Materials and processes used to model packaging for a roll of paper towels comes from Ecoinvent v2.2 (ecoinvent Centre, 2010). This includes folding boxboard for the paper towel core and polyethylene (LLDPE) for the plastic wrapper of the roll. P&G provided the mass of the core and wrapper. No secondary and transport packaging materials included. The contribution of packaging to the results is low (<1%) it is not included in the tables and graphs of the analysis.

Distribution, Use, and End-of-Life

The Bounty towels made at the two production facilities are distributed by tractor-trailer truck to a mix of distribution centers, clubs stores, and retailers across North America. Average distances range from 300 to 500 miles one-way and a load factor of 0.85 (representing trucks at 50% load capacity to pick up goods and at 100% capacity delivering goods) was assumed. No burdens were allocated to the retailer or the consumer to store, display, or use the product. The product does not contain chemicals that volatilize or leach, so there were no emissions to report during the use phase.

The roll of paper towels was assumed to be used for common household purposes, disposed, and hauled off with other household garbage to either a Municipal Solid Waste (MSW) landfill or incineration facility. US national average end of life treatment for the 2011 study year (US EPA, 2014) statistics provided the mix of landfill and incineration. Since used towels cannot be recycled because the fibers are two short to be commercially viable as a paper stock, the following equation estimates the percentage of towel waste to landfill:

where la is the reported average percentage of US MSW to landfill and ra is the reported average percentage recovered via recycling. The remaining percentage was assumed to be incinerated. GHG emissions and energy use related to end-of-life treatment were modeled with the EPA Waste Reduction Model (WARM) (US EPA, 2015). The openLCA database of WARM (Ingwersen et al, 2015) was used instead of the EXCEL® version of WARM to align WARM modelling choices of biogenic carbon with the approach taken in this study. In the model, each material was treated independently (towel, core, and wrapper). As the specific materials are not available in WARM, proxy

iiaiciiais wcic v^iiui>cn iui iiiulicii

IVIIACU

jjajjc

llllldl 11 >

1 C5IUC

ntial)' represented the towel, 'corrugated cardboard' represented the core, and 'LDPE' the plastic film wrapping. WARM by default accounts for "carbon storage" in the landfill, which is the total amount of carbon in the product remaining after partial decay. The total C stored is then converted to CO2-equivalents and subtracted from the total CO2-eq emissions to report GHG emissions, which is equivalent to the CO2 originally sequestered from the atmosphere by the biogenic source minus the C-equivalent that decays in the landfill. Since the CO2 originally sequestered was accounted for the forestry stage of the LCA model, the carbon storage in CO2 equivalents was not subtracted from the total CO2 emissions, to avoid double-counting. WARM by default does not track biogenic CO2 emissions from landfilling or combustion processes. These were then added in the model. The C content of the final product was assumed to be the same as C content of biomass. The percent of material combusted to CO2 was 98%, the same percentage used in combustion of other materials in WARM. Because landfill gas is on average 50% CO2 and 50% CH4 (Ingwersen et al, 2015), the amount of CO2 emission was set to match the amount of CH4 emissions. National average conditions and other default choices for WARM were used for model parameters.

Background Data

For all processes, production of generic chemicals, industrial water and wastewater treatment, and fuels other than petroleum are represented by data from Ecoinvent v2.2 (ecoinvent Centre, 2010). Data for petroleum fuels was taken from the inventories described in Sengupta et al. (Sengupta et al., 2015b). Data on crude oil and natural gas extraction as well as data from general forestry operations were taken from the USLCI Database (NREL, 2013).

Data quality

Data used in this study scores high for quality based on indicators like completeness, representativeness, consistency, reproducibility, data sources, technology coverage, precision, geographical coverage, time-related coverage, and uncertainty (ISO 2006b). The scoring approach is based on Weidema and Wesnaes, 1996; a semi-qualitative matrix pedigree method. The indicators of completeness, time-related coverage (temporal correlation), geographical coverage (geographical correlation), and technology coverage (technical correlation) matches the indicators in ISO 14044. Sample size compliments the completeness indicator. Temporal, geographical, and technical correlation describes representativeness. Consistency, reproducibility, and data sources are discussed throughout this section, Materials and Methods. Uncertainty is addressed with a sensitivity analysis. See Table A3 for the various data quality scores.

Sensitivity Analysis

Four types of scenarios were analyzed to understand the importance of new methods and datasets, as well as to evaluate inherent data uncertainties.

One scenario was developed to understand the life cycle result differences brought about by the new IPSA approach. In this scenario, a facility-level mass allocation approach was used to estimate the paper facility inventory to contrast with the new IPSA line level estimations.

Two scenarios were created explicitly to test how the specific datasets for pulp an

compared with national average data. In the first of these scenarios, US national average pulp data from 2011 was used to create a modified pulp LCI reflective of national conditions. Data from this scenario were taken from a dataset developed by the USEPA and others to support the Universal Industrial Sectors Integrated Solutions Model for the pulp and paper sector (USEPA, 2014; Modak et al., 2015). The national average energy inputs, and GHG and criteria pollutant emissions developed from these data, are presented in the Appendix Table A5. Other input process names and inputs were set to be identical to the existing pulp datasets to hold other factors constant. In the second scenario, the use of the new regionalized electricity LCI in the paper facility was replaced with the equivalent Ecoinvent 2.2 process for average US electricity at an industrial facility, which is 'electricity, medium voltage, at grid/US'.

Additional scenarios were developed to understand the importance of specific data accuracy on results; we systematically altered one or more of the key data points. Following initial model runs, key data points with inherent uncertainty were identified of particular potential importance to model results. Primary data provided by paper making facilities were of high quality in all aspects (ISO, 2006). Data of lesser quality and therefore with less certainty included data representing forestry and pulp operations (e.g. forestry yield, water use), product distribution, and end-of-life. For these points, values were doubled or halved. Ten scenarios were developed and applied to each line for a total of 20 scenarios run in the sensitivity analysis. New functionality was developed in openLCA 1.4 in collaboration with GreenDelta in order to conduct the sensitivity analyses. Within an openLCA 1.4 "Project," which is where different product systems can be compared, functionality to track variants of one or more systems were added. "Variants" were created with many variations of the baseline product system with a single change in one of the aforementioned key variables. Additional "variants" were created that used different product systems (or models) where unique unit processes has been substituted in the process network. This approach was taken to model the alternative facility allocation approach, and the scenarios with the substitution of the datasets, including the use of national average pulp and national average electricity datasets in place of the specific datasets used in the baseline case. The LCA results were then calculated simultaneously for each of these system variants.

2.2. Impact Assessment

Impact categories were chosen based on known impacts of concern, ability to contextualize impacts, availability of data to accurately represent potential impacts, and appropriateness of available impact methodologies. The selected categories are presented in Table 3 along with indicators to represent them. Impact indicators at the inventory analysis, midpoint, and endpoint are chosen along with normalization factors when available. Indicators from TRACI 2.1 (Bare, 2012) and ReCiPe 1.08 (Goedkoop et al., 2012) were chosen to represent potential environmental and human health effects at the midpoint level. TRACI models for these impacts are based on US conditions, where the majority of life cycle activities occur. ReCiPe was used to provide endpoint indicators for the same and additional impact categories and provide external normalization using global normalization factors (Sleeswijk et al., 2008; Van Hoof et al., 2013). Normalization is an optional approach in the ISO 14044 standards (ISO 2006) that can be used to view indicator scores in respect to a comparable reference point to aid

subjective weighting values; this interpretation implicitly implies an equal weighting value for all impact categories in the study. As a global consumer products company serving consumers with the variety of perspectives on which environmental issues are most critical, this approach is used as a way to narrow the indicators that are considered for further analysis, but not used as a reason for suggesting that other indicators are not important (Van Hoof et al. 2013). Where midpoint factors for resource depletion are not well developed in TRACI, methods were adopted from ReCiPe as well. The ReCiPe methodology is used in application to other P&G products and provides consistency across applications of LCA to different P&G products. The use of similar impact indicators from multiple LCIA methodologies provides a sensitivity check to help understand how results might differ across impact methods. Neither TRACI nor ReCiPe include indicators of energy use or solid waste generation. The Swiss Center for Life Cycle Inventories method for non-renewable energy demand was used (Frischknecht et al., 2007). To track water consumption, all evaporative losses as defined by the Water Footprint Network as blue water were tracked and aggregated by volume (WFN, 2009). Human health and ecotoxicity indicators were not used in this study due to lack of high quality data on toxics release to air and water across the entire life cycle, and due to manufacturing and consumer related releases that undergo detailed risk assessment, which is separate from LCA modeling.

In the climate change category for both TRACI and RECIPE methods, CO? uptake was assigned a C09-eq. value and -1 and biogenic CO? emissions were assigned a C09-eq of 1 so that both biogenic uptake and emissions were characterized to be consistent with the adjustments to the life cycle inventory. This results in the calculation of net global warming potential which can be represented in the following equation:

GWPnet = GWPgross + GWPuptake [2]

where GWPgross is the global warming potential of all greenhouse gas emissions regardless of fossil or biogenic origin, and GWPuptake is the global warming potential of plant uptake, which is always a negative value.

Table 2. Impact indicators used in this study, organized according to Bare and Gloria (Bare and Gloria, 2008). Impact categories names are those described in TRACI 2.1, except for categories which did not exist there, in which case names from ReCiPe or the other referenced methods are used.

Area of Protection Impact Category Material Flux Indicator Units Midpoint Indicator Units Endpoint Indicators, Units b Normalized? c

Human Health Effects Particulate matter kg PM2.5eq a, kg PM10eq b DALY Yes

Ozone depletion kg CFC11-eq a' b DALY Yes

Smog formation kg O3-eq a, kg NMVOC-eq b DALY Yes

Global climate kg CO2-eq a' b DALY, species.yrg Yes

Ionizing radiation kg U235-eq b DALY Yes

Natural Resources Fossil fuel depletion kg oil-eq b $ Yes

Metal depletion kg Fe-eq b $ Yes

Water Consumption/Depletion m3 d No

Cumulative Energy Demand, Non-renewable MJ e No

Land occupation/transformation 2* f m *yr Agricultural land, m2*yr b Urban land, m2*yr b Land transformation, m2 b Agricultural land, species.yr Urban land, species.yr Land transformation, species.yr Yes Yes Yes

Environmental Quality Freshwater eutrophication kg N-eq a, kg P-eq b species.yr Yes

Marine eutrophication kg N-eq b species.yr Yes

Acidification kg SO2-eq a' b species.yr Yes

a TRACI 2.1 (Bare, 2012)

b ReCiPe 1.08 with Hierarchist Perspective (Goedkoop et al., 2012)

c Normalization factor used were for Endpoint normalization factors in world per person impact d Water consumption is non-rainwater (bluewater) evaporative losses e (Frischknecht et al., 2007) f Total land occupation

g Global climate is normalized to both human health and environmental quality endpoints

3. Results

Figure 3 shows results of the LCA for 1 roll of paper towels made at Albany and Box Elder as a fraction of average global person consumption estimates using the ReCiPe endpoint normalization values. Fossil fuel depletion across the full paper towel life cycle is the largest potential indicator of environmental impact. The fossil fuel depletion score is one order of magnitude (~10x) greater than the next significant indicators, human health effects of climate change, agricultural land occupation, and human health effects of particulate matter formation, which in turn are an order of magnitude greater than the remaining indicator scores. Although use of normalization factors adds additional uncertainty to results (Van Hoof et al. 2013; Benini and Sala 2015; Weidema 2015), the profound differences increase the likelihood that are these are the impact categories with relatively higher contributions to average per person global impacts. It should also be noted that the differences between indicators are more substantial than the differences between Albany and Box Elder in the context of a global average consumer.

7.5E-04

6.5E-04

■s 5.5E-04

4.5E-04

S 3.5E-04

g> 2.5E-04

g 1.5E-04

5.0E-05

-5.0E-05

I Albany I Box Elder

^ J* ^ if if if if

s / * s ^

Figure 3. Normalized results of producing 1 roll of Bounty" paper towels at Albany and Box Elder.

Figure 4 is a comparison of results for the most significant impact categories for the production of Bounty at Albany and Box Elder based on the endpoint normalization, along with water consumption. The impact scores are internally normalized to reflect 100% of the highest score. Water consumption is an additional indicator not available for normalizing to annual world emissions, but is relevant for products that are made with pulp, as their processing requires significant amounts of water. Box Elder has lesser impacts for fossil fuel depletion, climate change, and land occupation, whereas Albany has a lesser impact on particulate matter formation and water consumption. The differences range from 7%

for fossil fuel depletion to 54% for water consumption. These differences are explained in the contribution analysis.

Fossil Depletion Climate Change Agricultural Land Particulate Water

Use Matter Consumption

Albany Box Elder

Figure 4. Comparison of results for selected indicators calculated for the production of Bounty® at Albany and Box Elder. The impact scores are internally normalized to reflect 100% of the highest score.

A contribution analysis enables a better understanding of the most significant life cycle stages and drivers for each impact category, and provides a "hotspot analysis" to focus future research and development. Based on early analysis, life cycle groupings that showed distribution of impacts by upstream contributors and downstream stages were identified. Upstream impacts were largely distributed across the electricity, fuel, and pulp supply chains for Bounty production. Direct impacts of the Bounty production facility were important to distinguish, as well as impacts related to the downstream stages of distribution and end-of-life.

Figure 5A shows the contribution analysis for fossil fuel depletion, the indicator most significant from the normalized comparison, for the paper towels made at each facility. Pulp purchased to make the paper towels is the largest contributor (49%). Some of the energy for making pulp is derived from nonfossil resources (~5%). The pulp combinations used at Albany contribute slightly more than the pulp materials at Box Elder (0.19 kg oil eq vs. 0.16, respectively). The residual wood from debarking and chipping are used as fuel for the pulp production. However, the wood milling process (making of logs, debarking, and chipping) uses residual fuel oil boilers (Wagner et al., 2009), which drives the fossil fuel depletion indicator in this case. Facility electricity and fuels are the next leading contributors to fossil fuel depletion following pulp production. A larger share of electricity purchased by the older Albany comes from fossil fuel sources because of the local electricity grid. The newer Box Elder has an overall smaller fuel energy requirement, for which it uses natural gas. Although Albany uses biomass from milling wastes (a non-fossil fuel source) for a percentage of its energy needs, the

residues from debarking and chipping required fossil fuels for their harvesting and processing. Distribution of the product to wholesale and retailers requires long-distance trucking, which contributes approximately 6% to fossil fuel depletion.

Global climate change was modeled to include both emissions of GHGs as well as sequestration of atmospheric CO2 by trees used for pulp and fuel production. Significant contributors to emissions include pulp production, facility electricity, facility production lines, and end-of-life consumer disposal (Figure 5B). Pulp production contributed the greatest share to GHG emissions. Pulp production, however, results in significant reductions due to carbon sequestration during tree production. The net global warming potential from pulp is slightly negative (-0.15 to -0.25 kg CO2-eq), as the sequestration slightly outweigh the emissions from the supply chains. The mix of tree species was different at each facility. Tree species production for pulp purchased by Box Elder provided more carbon sequestration, but Albany showed more carbon sequestration overall when the biomass fuels at the facility were included. Facility electricity is the next most contributing life cycle stage, dominated by electricity from coal. Production Line and disposal emissions are relatively small as the third contributor, and Box Elder has reduced its lines' emissions more than Albany (0.16 kg CO2-eq vs. 0.22 kg CO2-eq). Disposed Bounty is assumed to degrade anaerobically in landfills, contributing to some landfill methane emissions.

Agricultural land occupation is dominated by the Pulp life cycle stage (Figure 5C). The Pulp

combination used at the older Albany has slightly greater land occupation than Pulp used at Box Elder

(2.3 m .yr vs. 2.2 m .yr, respectively). This life cycle indicator is influenced by the differences in pulp suppliers; which include differing tree species, climate, time to yield, and operations. Details on pulp supply and mix purchased by the facilities cannot be disclosed due to legal agreements on data confidentiality.

The particulate matter health effects indicator results were different between facilities, as shown in Figure 5D. Particulate emissions < 2.5 um for the newer facility are associated with the production line and other activities, while the older facility had sulfur dioxide and particulate emissions < 2.5 um related to pulp production and electricity purchased for the manufacturing. A portion of both facilities' electricity came from coal and biomass, which contributes most to respiratory effects. The pulp purchased by Albany are responsible for more potential respiratory effect than that purchased by Box Elder.

Other indicators that have no normalization factors, but are relevant for product systems like paper towels, are analyzed. Those indicators include Cumulative Energy Demand and Water Consumption. Cumulative Energy Demand (Figure 5E) reflects a nearly identical contribution analysis profile as the Fossil Fuel Depletion chart (Figure 5A), in that the largest contributing phases are pulp followed by electricity and facility fuel.

El other

■ wood production for pulp/fuel

■ electricity from nat. gas

■ landfilling/incineration

■ electricity from coal □ pulp production

Electricity

_ ..........

Albany Box Elder Albany Box Elder Albany Box Elder Albany Box Elder

Fuel Pulp Pulp Transport Bounty prod, line Distribution Disposal

□ other

□ wood production for pulp/fuel

Albany Box Elder Albany Box Elder Albany Box Elder Albany Box Elder Albany Box Elder

Electricity Fuel Pulp Pulp Transport Bounty prod. Distri- Disposal

line bution

E3 other

s nat. gas processing

□ pulp production

S electricity from biomass

□ electricity from coal

Albany Box Elder Albany Box Elder Albany Box Elder Albany Box Elder Albany Box Elder

Electricity Fuel Pulp Pulp Transport Bounty prod. Distri- Disposal

line bution

Albany Box Elder Electricity

Albany Box Elder Fuel

Albany Box Elder Pulp

Albany Box Elder Pulp Transport

Albany Box Elder

Bounty prod, line

Distribution

Disposal

Figure 5. Comparison of contributors to (A) fossil fuel depletion (kg oil-eq), (B) climate change (kg CO2-eq), (C) agricultural land occupation (m2), (D) particulate matter formation (kg PM2.5-eq), E) cumulative energy demand (MJ primary energy), and (F) water consumption (m3 of water). Contributions are presented by process and further broken down by key life cycle stage components. 'Electricity' represents electricity for Bounty production; 'Fuel' is fuel for Bounty production; 'Pulp' is pulp production; 'Distribution' is distribution of Bounty to retailers; and 'disposal' is disposal of Bounty by the consumer. For these latter two stages, impacts apply equally to both lines, otherwise they are independent.

For water consumption (Figure 5F) the dominant contributor varies by facility. The newer Box Elder facility is more water efficient during production; however, electricity purchases by the facility dominate water consumption for Box Elder by 64%. Box Elder purchases electricity with a high contribution of hydropower in the grid, resulting in large evaporative water loss in reservoirs used to generate hydropower, which dominated the facility's life cycle water consumption. This water consumption is three times greater than direct water consumption at the production stage, which contributed 19% of total water consumption, just 3% greater than the contribution of pulp for this facility. For the older Albany facility, the production line requires more water compared to all other life cycle stages (47% of total), followed by pulp (29%) and electricity (24%). Pulp production is the next most contributing component. The pulp mix used at Box Elder is slightly more water intensive than the pulp mix used at Albany.

4. Sensitivity Analysis

IPSA vs traditional facility level allocations

Results of the scenario to understand if/which differences are brought about by using the new IPSA approach vs. a more traditional facility-level mass allocation approach are shown in Table 3. Results show that the IPSA method produces line-specific input and emissions estimates that differ from facility averages. For Albany, the IPSA-based quantities are less than the traditional mass allocation approach by 14% for most inputs, and vary from 6 to 61% less for the emissions. This can be interpreted that the Albany production system for producing the paper towel during the time period of interest was less input-intensive and emitted less than the average line in the facility. On the other hand, the IPSA-based quantities for Box Elder are about the same as the average for the facility (0-2%+). There is less variation in Box Elder from the facility average because there was a single papermaking production line online during the sampling period, so the variation is only explained by the differences in the paper conversion line.

Table 3. Relative Paper Facility Flow Amounts from IPSA Procedure as % of Facility Level Mass Allocation.

Albany Box Elder

Inputs

Pulp 86% 102%

Chemicals (average) 86% 102%

Fuels (average) 86% 100%

Water 86% 100%

Electricity 89% 100%

Outputs

PM10 94% 102%

PM2.5 93% \ 102%

SO2 86% 102%

NOx 39% 102%

VOC 84% 102%

CO 61% 102%

Lead 86% 102%

NH3 87% 102%

CO2 68% 102%

Wastewater 86% 102%

Water loss 86% 102%

Pulp and regional electricity scenarios

Using national average pulp data on energy use and emissions resulted in the largest differences of any scenario. The input data for national average pulp is more energy and criteria pollutant emissionintensive than the pulps used by the facilities, but emits less GHGs. Due to this apparent inconsistency in these results, a consistency check was performed to compare the GHG emissions from these two datasets. Pulp mills use a number of energy sources including many internally recycled biogenic

sources, and there might be differences in accounting procedures for GHGs as a result that could not be reconciled. The use of an alternate proxy material for paper towels in the WARM model resulted in a significant decrease in disposal phase and full life cycle emissions.

Key data scenarios

The other data points that could have significant influence on results include the water loss in pulp production and the distribution distance; neither had much influence over the life cycle results. The forest yield changes do impact land use since it is dominated by the forestry, but otherwise yield does not affect life cycle impacts.

Tables 4 and 5 show the results of the sensitivity analyses for Albany and Box Elder, respectively, for the impact categories of most significance. For Albany, the IPSA procedure had a significant impact on results, since the line showed different performance than other lines in that facility when compared to the facility-level mass allocation. Having accurate numbers for forest yield and supply chain specific pulp data would also influence results for that facility. Results for Box Elder were similar, expect the IPSA method was not as significant; this is due to only having the one line running so the facility level data already reflected just that line, not a line average.

Table 4. Relative changes to full life cycle impact indicator results for Albany from the sensitivity analyses of allocation method, distribution distance, pulp, electricity, forest yield and water used in forestry.

Scenario Fossil Climate Ag. Land Particulate Energy Water

fuel depletion change Occupation Matter Demand Consumption

High forest yield 100% 100% 67% 100% 100% 100%

Low forest yield 100% 100% T 133% 100% 100% 100%

High water loss for NA NA NA NA NA 99%

Low water loss for NA NA NA NA NA 101%

National average pulp 166% 84% NA 113% 160% NA

vs supply chain

specific data

National average 98% 103% 100% 99% 99% 76%

electricity vs regional

Mass allocation vs 114% 120% 116% 114% 115% 115%

Long distribution 104% 104% 100% 102% 104% 100%

distance

Short distribution 98% 98% 100% 99% 98% 100%

distance

Alternate material for NA 82% NA NA 100% NA

Note: Bolded values indicate sensitivities of 10% or higher.

Table 5. Relative changes to full life cycle results for Box Elder from sensitivity analysis of allocation method, distribution distance, pulp, electricity, forest yield and water used in forestry.

Scenario Fossil fuel depletion Climate change Ag. Land Occupation Particulate Matter Energy Demand Water Consumption

High forest yield 100% 100% 67% 100% 100% 100%

Low forest yield 100% 100% 133% 100% 100% 100%

High water loss for pulp NA NA NA NA NA 92%

Low water loss for pulp NA NA NA NA NA 104%

National average pulp vs supply chain specific data 211% 85% NA 127% 201% NA

National average electricity vs regional 105% 125% 100% 111% 107% 36%

Mass allocation vs IPSA 99% 100% 98% 98% , 99% 99%

Long distribution distance 107% 106% 100% 103% 107% 100%

Short distribution distance 96% 97% 100% 99% 97% 100%

Alternate material for WARM NA 80% NA NA 100% NA

Note: Bolded values indicate sensitivities of 10% or higher.

5. Discussion and conclusions

LCA studies can vary in quality due to data, methodologies, allocations, assumptions, modeling choices, etc. (Bousquin et al., 2010; Subramanian et al., 2011). Extra care is needed to use the best data and methodologies possible so that sustainability goals and strategies are based on robust science. This study used innovative and technically strong approaches to fortifying data (pulp, manufacturing, electricity), allocation (IPSA), and modeling (openLCA, WARM). The iterative nature of the LCA resulted in a recognition of the need to improve the quality of pulp and electricity LCI, making the results for these important contributors more accurate. Pulp production LCI specific to the most important pulp sources for Bounty production were developed, providing results that appear to differ significantly from use of average US pulp data. New electricity LCI data were developed, which better reflects the technologies currently in use in the US. This is important in the paper towel life cycle, particularly for facilities that draw from a regional electricity grid with a much different mix than the national average. The data allocation at the complex, multi-line manufacturing facility is superior to previous methods of simple mass or economic allocation, since actual line metrics were utilized in the new IPSA methodology to avoid arbitrary allocation. This is particularly important in providing linelevel inventory and distinguishing many lines of varying performance at a facility. Additional functionality was added to openLCA software that provided a more straightforward and consistent means of performing sensitivity analysis for LCA that will be of benefit to the global LCA community of practitioners. OpenLCA 1.4 proved to be effective for performing and managing a detailed LCA study of a product from a major manufacturer. This study is foundational to the exploration of other integrated sustainability metrics that incorporate the environmental estimates from this study, with financial and social data.

The conclusions of this study are clear, based on this work and other LCAs on paper towels (Montalbo et al, 2011; Madsen, 2007; Joseph et al, 2015), that making the product drives much of the relevant impacts on the environment. Energy requirements at the plants are high and draw from fuels and electricity grids that are predominantly fossil-based. However, this study also revealed that use of facility-level metrics alone to drive facility-level changes are insufficient to address all significant life cycle impacts. Analyzing beyond the papermaking and converting plants, pulp supply contributes greatly to the life cycle impact of paper towels as well. Other life cycle stages, including disposal, are not insignificant and therefore metrics that capture the full life cycle context are indeed needed to provide the proper context to inform sustainability-related decisions and identify potential changes that might be made beyond the facility.

In this study, both the manufacturing facility's life cycle impacts are based on the same unit, 1 roll of Bounty paper towels; yet the analysis still identified differences in environmental impacts driven by different technological processes and locations. Box Elder is a newer plant with a state of the art platform, and incorporates design features to improve efficiency. Bounty produced at Box Elder has potentially a lesser impact on global warming potential (-9%) and fossil fuel depletion (-8%). Agricultural land occupation is lower for Box Elder (-16%), which is more a reflection of pulp supply mix than Box Elder's processing and operations. Box Elder results in more life cycle water consumption (+54%), due not to plant operations but electricity production in the region of the facility, and more potential health effects of particulate matter emissions (+14%) than Albany, which is an older site.

Within the scope of this study, fossil fuel depletion is the most relevant impact indicator, and looking at both direct and indirect means of reducing fossil fuel usage should be the highest priority for reducing overall impacts. Electricity production is identified as being significant for many of the impact categories including fossil fuel depletion. Both facilities use nearly identical electricity amounts per unit output paper towel. Differences in the specific energy mix in each region were significant. The electricity mix supporting Albany, utilizes a grid powered by more than 50% coal, which drove the calculated impacts (Appendix Table A1). Subsequently, both facilities, especially Albany, would benefit from less dependency on coal power or through otherwise acquiring more electricity from less fossil-fuel intensive sources. Reductions in the dependency of grid electricity would reduce impacts in fossil fuel depletion, respiratory effects, and global warming potential. However, increasing the use of wood residues, which reduces the need for grid electricity, will increase agricultural land occupation. The Albany plant recently announced development of an up to 50-megawatt biomass plant on-site. Because Albany is one of P&G's largest U.S. facilities, the project will significantly increase P&G's use of renewable energy, helping move the company closer to its 2020 goal of obtaining 30% of its total energy from renewable sources (Procter & Gamble, 2015). P&G will need to stay diligent in its commitment to sustainable sourcing so that expanding the land area needed to support energy needs at the facilities will not impair forest value and services. P&G's wood procurement policy (Procter & Gamble, 2014) addresses sustainable forest management, certifications, conversions, and its efficient use of resources. This is designed to ensure responsible long-term supply, enabling reliable quality and availability of paper products.

The fossil fuel energy use information, along with the global climate change estimates, could be used to encourage pulp suppliers to continue to reduce their energy use and greenhouse gas emissions during production. The results showed the pulp combination used at Box Elder can provide more carbon sequestration, although the pulp types used at Albany had higher carbon sequestration overall when the biomass fuels at the facility were included in the modelling scenarios. When including the potential for carbon sequestration as criteria for product design, pulp production processes could deliver the most significant reductions in the climate change indicator, along with fuel purchasing (of biomass for energy) at the Albany.

For water consumption, this study helps to recognize that electricity sources, even those from renewables, can lead to impacts that dominate the life cycle; such as the hydropower-associated water losses that dominate water use for Box Elder. Hydropower is an important contributor to the power mix in that region. Operational water consumption factors for aggregate US in-stream and reservoir hydropower for a median of 4,491 gallons/MWh. Other renewable electricity sources have water consumption factors ranging from 0-1,000 gallons/MWh (Macknick et al., 2012; Mekonnen and Hoekstra, 2012). Continuous improvement programs designed to use less water and recycle more can yield life cycle benefits, provided that burden-shifting is avoided. Further this study illustrate that technology improvements at the older Albany can reduce water consumption enough to match the newer lines at Box Elder, which would result in less water impacts from product production. Water reduction or recycling strategies at the pulp and paper making facilities would make the most meaningful improvements to this indicator.

P&G has a long history of developing LCA methods and studies, and using the results to set practical and meaningful sustainability goals. Previous studies identified the importance of reducing resources and emissions from paper plant operations. This more detailed study has four main findings with practical application to paper products and other company product families. First, although the IPSA method has more steps than the traditional facility-level allocation method, it is not difficult to use and delivers more accurate input data for LCA. P&G will apply the IPSA method for inputs to other product LCAs in the future. Other manufacturers and LCA practitioners may be persuaded to use IPSA as well, as the algorithms are published and described. Secondly, by applying a normalization step, the study identifies the most important impact indicators that P&G should track for this product to enable more sustainable production. The top 6 indicators will continue to be monitored and improved, but focusing on energy types and use are critical. Third, defining the meaningful differences between plants, driven by established vs. new technologies and/or location, is also important. For example, Figs. 4 and 5 show the importance of understanding particulate emissions at Box Elder, and determining whether the higher water consumption due to reservoir evaporation at hydroelectric dams is something that P&G deems of relevant or minimal concern. Fourth, by conducting a series of sensitivity analysis coupled with detailed contribution analysis, P&G can understand where better data and methods will improve model accuracy, as well as estimate how changes in facility operations, supplier activities, or otherwise product characteristics can change life cycle impacts.

Acknowledgments

This manuscript was developed through a Cooperative Research and Development Agreement

(CRADA, No. 683-12) between the U.S. Environmental Protection Agency National Risk

Management Research Laboratory and The Procter & Gamble Company. The contributions of Drs.

Sengupta and Lee were supported by an appointment to the Postdoctoral Research Program at the U.S.

EPA, National Risk Management Research Laboratory, administered by the Oak Ridge Institute for

Science and Education through an Interagency Agreement between the U.S. Department of Energy and

the U.S. EPA.

Conflict of Interest

The authors declare no conflict of interest.

Disclaimer

This article does not reflect the endorsement or opinion of the US Environmental Protection

Agency.

References

Bare, J., 2012. Tool for the Reduction and Assessment of Chemical and other Environmental Impacts (TRACI): TRACI Version 2.1 - User's Manual. US EPA National Risk Management Laboratory, Cincinnati, OH.

Bare, J.C., 2010. Life cycle impact assessment research developments and needs. Clean technologies and environmental policy 12, 341-351.

Bare, J.C., Gloria, T.P., 2008. Environmental impact assessment taxonomy providing comprehensive coverage of midpoints, endpoints, damages, and areas of protection. Journal of Cleaner Production 16, 1021-1035.

Benini, L., Sala S., 2015. Uncertainty and sensitivity analysis of normalization factors to methodological assumptions. Int J Life Cycle Ass 21, 224-236.

Binkley, D., 2014. Brazil Eucalyptus Potential Productivity (BEPP).

Boguski, T.K., 2010. Life cycle carbon footprint of the National Geographic magazine. Int J Life Cycle Ass 15, 635-643.

BSI, 2011. PAS 2050: 2011 Specification for the assessment of the life cycle greenhouse gas emissions of goods and services. British Standards Institute.

BTS, 2009. 2007 Economic Census: Transportation Commodity Flow Survey. U.S. Department of Transportation, Research and Innovative Technology Administration, Bureau of Transportation Statistics and U.S. Department of Commerce, U.S. Census Bureau, Washington, DC.

Cai, H., Wang, M., Elgowainy, A., Han, J., 2012. Updated greenhouse gas and criteria air pollutant emission factors and their probability distribution functions for electricity generating units. Argonne National Laboratory (ANL).

Cochran, P.H., Dahms, W.G., 2000. Growth of Lodgepole Pine Thinned to Various Densities on Two Sites With Differing Productivities in Central Oregon. United States Forestry Service, United States Department of Agriculture.

Curran, M.A., 1996. Environmental life-cycle assessment. Int J Life Cycle Assess 1, 179-179.

Curran, M.A., 2000. Life Cycle Assessment: An international experience. Environmental Progress 19, 65-71.

EC, 2013. ANNEX II : PRODUCT ENVIRONMENTAL FOOTPRINT (PEF) GUIDE to the COMMISSION RECOMMENDATION on the use of common methods to measure and

communicate the life cycle environmental performance of products and organisations. European Comission, Brussels.

ecoinvent Centre, 2010. ecoinvent data v2.2. Swiss Centre for Life Cycle Inventories, Dubendorf, CH.

EIA, U., 2015. International Energy Statistics US Energy Information Administration.

EPA, U., 2010. Avilable and emergy technologies for reducing GHGs from the pulp and paper manufacturing industry. Office of Air and Radiation, Washington, DC.

EPA, U., 2014. Emissions & Generation Resource Integrated Database (eGRID), Year 2010 Data. Ninth Edition.

Fava, J., Denison, R., Jones, B., Curran, M., Vigon, B., Selke, S., Barnum, J., 1991. A technical framework for life-cycle assessment, Pensacola, Florida.

Fava, J.A., 1991. A technical framework for life-cycle assessments. Society of Environmental Toxicology and Chemistry and SETAC Foundation for Environmental Education.

Finnveden, G., Hofstetter, P., Bare, J., Basson, L., Ciroth, A., Mettier, T., Seppala, J., Johansson, J., Norris, G., Volkwein, S., 2002. Normalization, grouping and weighting in life cycle impact assessment. Lifecycle impacts assessment striving towards best practice. SETAC. Pensacola Fl. USA.

FP Innovations. 2015. Product Category Rules for Market Pulp. Vancouver, British Columbia.

Frischknecht, R., Jungbluth, N., Althaus, H.-J., Doka, G., Dones, R., Heck, T., Hellweg, S., Hischier, R., Nemecek, T., Rebitzer, G., Spielmann, M., Wernet, G., 2007. Overview and methodology, Ecoinvent report No. 1. Swiss Centre for Life Cycle Inventories, Dubendorf.

Gaudreault, C., Samson, R., Stuart, P.R. 2010. Energy decision making in a pulp and paper mill: selection of LCA system boundary. International Journal Life Cycle Assessment 15: 198-211.

GHG Protocol, 2011. Product Life Cycle Accounting and Reporting Standard. World Resources Institute, World Business Council for Sustainable Development, Washington and Geneva.

Goedkoop, M., Heijungs, R., Huijbregts, M., Schryver, A.D., Struijs, J., Zelm, R.v., 2012. ReCiPe 2008: A life cycle impact assessment method which comprises harmonised category indicators at the midpoint and the endpoint level.

Gooch, J., Mack, R., 2012. Press Release - P&G and US EPA Sign Research Agreement to Help Company Attain Sustainability Goals.

Ingwersen, W.W., Ceja, M., Weisbrod, A.V., Cabezas, H., Demeke, B., Eason, T., Smith, R.L., Sengupta, D., Zanoli, E., Gausman, M., Lee, S.-J., Ma, X.C., Weber, B., Alvarez, M., Bare, J.C., Abraham, J., Ruiz-Mercado, G.J., Gonzalez, M.A., 2016. Evaluating consumer product life cycle sustainability with integrated metrics: A Bounty® Case Study. Industrial & Engineering Chemistry Research. DOI: 10.1021/acs.iecr.5b03743

Ingwersen, W.W., Rodriguez, C., Greve, S., Srocka, M., Ciroth, A., Meyer, D., 2015. WAste Reduction Model (WARM) for openLCA, 29th Meeting of the International Expert Group for

LCA for Integrated Waste Management, Washington, DC.

Ingwersen, W., Cabezas, H., Weisbrod, A.V., Eason, T., Demeke, B., Ma, X.C., Lee, S.-J., Hawkins, T., Bare, J.C., Ceja, M., 2014. Integrated Metrics for Improving the Life Cycle Approach to Assessing Product System Sustainability. Sustainability 6, 1386-1413.

Ingwersen, W., Curran, M., Gonzalez, M., Hawkins, T., 2012. Using screening level environmental life cycle assessment to aid decision making: A case study of a college annual report. International Journal of Sustainability in Higher Education 13, 6-18.

Ingwersen, W., Subramanian, V., 2013. Guidance for product category rule development: process, outcome, and next steps. Int J Life Cycle Ass 19, 532-537.

Ingwersen, W., Weisbrod, A.V., Gausman, M., Lee, S.-J., Sengupta, D., Smith, R., Demeke, B., Young, D., Abraham, J., Hawkins, T., Zanoli, E., Adams, D., Ruiz-Mercado, G., Gonzalez, M., Ma, C., Eason, T., Cabezas, H., Bare, J., Hart, K., Ceja, M., 2013. LCA models at the core of the sustainable supply chain toolkit: A paper towel case study, International Symposium on Sustainable Systems and Technology (ISSST) 2013, Cincinnati, OH.

ISO, 2006. 14044: Environmental management -- Life cycle assessment -- Requirements and guidelines. International Standards Organization, Geneva.

Jolliet, O., Frischknecht, R., Bare, J., Boulay, A.-M., Bulle, C., Fantke, P., Gheewala, S., Hauschild, M., Itsubo, N., Margni, M., 2014a. Global guidance on environmental life cycle impact assessment indicators: findings of the scoping phase. Int J Life Cycle Assess 19, 962-967.

Jolliet, O., Frischknecht, R., Bare, J., Boulay, A.-M., Bulle, C., Fantke, P., Hauschild, M., Margni, M., McKone, T., Canals, L.M.y., Postuma, L., Prado, V., Ridoutt, B., Sonneman, G., Rosenbaum, R., Struis, J., Zelm, R.v., Vigon, B., Weisbrod, A., 2014b. Global Guidance on Environmental Life Cycle Impact Assessment Indicators: Findings of the Glasgow scoping workshop. International Journal of Life Cycle Assessment.

Jolliet, O., Müller-Wenk, R., Bare, J., Brent, A., Goedkoop, M., Heijungs, R., Itsubo, N., Peña, C., Pennington, D., Potting, J., 2004a. The LCIA midpoint-damage framework of the UNEP/SETAC life cycle initiative. Int J Life Cycle Assess 9, 394-404.

Jolliet, O., Müller-Wenk, R., Bare, J., Brent, A., Goedkoop, M., Heijungs, R., Itsubo, N., Peña, C., Potting, J., Pennington, D., Rebitzer, G., Schenck, R., Stewart, M., Udo de Haes, H., Weidema, B., 2004b. The LCIA midpoint-damage framework of the UNEP-SETAC Life Cycle Initiative. International Journal of LCA 9, 394-404.

Joseph, T., Baah, K., Jahanfar, A., Dubey, B. 2015. A comparative life cycle assessment of conventional hand dryer and roll paper towel as hand drying methods. Science of the Total Environment 515-516: 109-117.

Kulak, M., Nemecek, T., Frossard, E., Gaillard, G. 2016. Eco-efficiancy improvement by using integrative design and life cycle assessment. The case study of alternative bread supply chain in France. Journal of Cleaner Production 112: 2452-2461.

Lee, S.-J., Ingwersen, W.W., Hawkins, T.R., 2015. Development of Open-source Life cycle Inventory Data for Electricity Generation in the United States, Including Water, Greenhouse Gas, and Land Use.

Macknick, J., Newmark, R., Heath, G., Hallett, K., 2012. Operational water consumption and withdrawal factors for electricity generating technologies: a review of existing literature. Environmental Research Letters 7, 045802.

Madsen, J. 2007. Life Cycle Assessment of Tissue Products Final Report. Environmental Resource Management.

Margni, M., Gloria, T., Bare, J., Seppälä, J., Steen, B., Struijs, J., Toffoletto, L., Jolliet, O., 2007. Guidance on how to move from current practice to recommended practice in Life Cycle Impact Assessment, in: UNEP/SETAC Life Cycle Initiative (Ed.). Task Force 1 of the UNEP/SETAC Life Cycle Initiative, Paris, France.

McDougall, F.R., White, P.R., Franke, M., Hindle, P., 2008. Integrated solid waste management: a life cycle inventory. John Wiley & Sons.

Mekonnen, M. M., Hoekstra, A. Y., 2012. The blue water footprint of electricity from hydropower. Hydrology and Earth System Sciences 16, 179-187.

Meyer, D.E., Upadhyayula, V.K., 2014. The use of life cycle tools to support decision making for sustainable nanotechnologies. Clean Technologies and Environmental Policy 16, 757-772.

Modak N, Spence K, Sood S, Rosati JA (2015) Building a Comprehensive Mill-Level Database for the Industrial Sectors Integrated Solutions (ISIS) Model of the U.S. Pulp and Paper Sector. PLoS ONE 10(3): e0120954. doi:10.1371/journal.pone.0120954

Montalbo, T., Gregory, J., Kirchain, R. 2011. Life Cycle Assessment of Hand Drying Systems. Materials Systems Laboratory, Massachusetts Institute of Technology.

Mutel, C.L., Pfister, S., Hellweg, S., 2011. GIS-Based Regionalized Life Cycle Assessment: How Big Is Small Enough? Methodology and Case Study of Electricity Generation. Environmental Science & Technology 46, 1096-1103.

NREL, 2013. US Life Cycle Inventory Database.

Procter & Gamble, 2015, Procter & Gamble and Constellation Announce One of Nation's Largest Biomass Renewable Energy Plants Scrap wood to create 100% of the steam needed to run one of P&G's largest paper manufacturing facilities Category: P&G Corporate Announcements

Thursday, February 12, 2015 9:00 am EST. http://news.pg.com/press-release/pg-corporate-announcements/procter-gamble-and-constellation-announce-one-nations-large), Last accessed 29 February 2016.

Procter & Gamble, 2014. P&G Wood Pulp Procurement Policy. https://www.pg.com/en_ANZ/sustainability/points-of-view/wood-pulp.shtml. Last accessed 29 Februar 2016.

Procter & Gamble, 2015. Sustainability Report.

Puettmann, M.E., Bergman, R., Hubbard, S., Johnson, L., Lippke, B., Oneil, E., Wagner, F.G., 2010. Cradle-to-gate life-cycle inventory of US wood products production: CORRIM Phase I and Phase II products. Wood and fiber science 42, 15-28.

Rodriguez, C., W. W. Ingwersen and A. Ciroth (2015). Life Cycle Inventory Database of the EPA WAste Reduction Model (WARM) for openLCA. Cincinnati, OH, US Environmental Protection Agency, Office of Research and Development, National Risk Management Research Laboratory, Life Cycle Assessment Research Center.

SAIC, Curran, M.A., 2006. Life-cycle assessment: principles and practice. National Risk Management Research Laboratory, Office of Research and Development, US Environmental Protection Agency.

Saouter, E., van Hoof, G., Feijtel, T.C., Owens, J.W., 2002. The effect of compact formulations on the environmental profile of Northern European granular laundry detergents Part II: Life Cycle assessment. Int J Life Cycle Assess 7, 27-38.

Schmidt, J.H., Thrane, M., Merciai, S., Dalgaard, R., 2015. Inventory of country specific electricity in LCA - consequential and attributional scenarios. Methodogy report v2. 2.-0 LCA consultants, Aalborg, Denmark.

Sengupta, D., Abraham, J.P., Ceja, M., Gonzalez, M.A., Ingwersen, W.W., Ruiz-Mercado, G.J., Smith, R.L., 2015a. Industrial Process System Assessment: Bridging Process Engineering and Life Cycle Assessment through Multiscale Modeling. Journal of Cleaner Production 90, 142-152.

Sengupta, D., Hawkins, T.R., Smith, R.L., 2015b. Using national inventories for estimating environmental impacts of products from industrial sectors: a case study of ethanol and gasoline. Int J Life Cycle Assess 20, 597-607.

Sleeswijk, A.W., van Oers, L.F., Guinée, J.B., Struijs, J., Huijbregts, M.A., 2008. Normalisation in product life cycle assessment: An LCA of the global and European economic systems in the year 2000. Science of the total environment 390, 227-240.

Subramanian, V., Inwersen, W., Hensler, C., Collie, H. Comparing PCRs from Different Programs: Learned Outcomes Towards Global Alignment. LCA XI, Chicago, October 4-6, 2011 Special Session: PCR Alignment

Todd, J.A., Curran, M.A., Weitz, K., Sharma, A., Vigon, B., Price, E., Norris, G., Eagan, P., Owens, W., Veroutis, A., 1999. Streamlined life-cycle assessment: a final report from the SETAC North America streamlined LCA workgroup. Society of Environmental Toxicology and Chemistry (SETAC) and SETAC Foundation for Environmental Education.

USEPA.2015.GreenhouseGas Reporting Program (GHGRP).

https://ghgreporting.epa.gov/ghg/login.do

US DOE, 2015. The Greenhouse Gases, Regulated Emissions, and Energy Use in Transportation (GREET) Model. US Department of Energy, Argonne National Laboratory Transportation Technology R&D Center, Argonne, IL.

US EPA, 2014. Municipal Solid Waste Generation, Recycling, and Disposal in the United States: Facts and Figures for 2012.

US EPA. 2014. Universal Industrial Sectors Integrated Solutions Model for the Pulp and Paper Manufacturing Industry - UISIS-PNP. Version 3.1.

US EPA. 2015. Inventory of U.S. Greenhouse Gas Emissions and Sinks: 1990-2013. EPA 430-R-15-004. Washington, DC.

US EPA, 2015. Waste Reduction Model (WARM), v13. US Environmental Protection Agency, Washington, DC.

USFS, 1983. Silvicultural Systems for the Major Forest Types of the United States. United States Forestry Service.

Van Hoof, G., Vieira, M., Gausman, M., Weisbrod, A., 2013. Indicator selection in life cycle assessment to enable decision making: issues and solutions. Int J Life Cycle Assess 18, 15681580.

Van Hoof, G., Weisbrod, A., Kruse, B., 2014. Assessment of Progressive Product Innovation on Key Environmental Indicators: Pampers® Baby Wipes from 2007-2013. Sustainability 6, 51295142.

Wagner, F.G., Puettmann, M.E., Johnson, L.R., 2009. Life-cycle inventory of Inland Northwest softwood lumber manufacturing. CORRIM Phase II Report Module B. Environmental performance measures for renewable building materials with alternatives for improved performance. Seattle, WA.

Weidema, B. P., 2015. Comparing three life cycle impact assessment methods from an endpoint perspective. Journal of Industrial Ecology 19(1), 20-26.

Weidema, B. P., Wesnaes, M. S. 1996 Data quality management for life cycle inventories - an example of using data quality indicators. Journal of Cleaner Production 4: 167-174.

Weisbrod, A., Loftus, L., 2012. A Case Study of the Practice of Sustainable Supply Chain Management. Life Cycle Assessment Handbook: A Guide for Environmentally Sustainable Products, 233-248.

Weisbrod, A.V., Van Hoof, G., 2012. LCA-measured environmental improvements in Pampers® diapers. Int J Life Cycle Assess 17, 145-153.

WFN, 2009. Water Footprint Manual. Water Footprint Network, Entschede, NL.

Young, D.M., Hawkins, T., Ingwersen, W., Lee, S.-J., Ruiz-Mercado, G., Sengupta, D., Smith, R.L., 2012. Designing sustainable supply chains. CHEMICAL ENGINEERING 29.

Appendix

Table A1. Electricity power mix for the regions in which the facilities were located in comparison with the US average in 2011 Source: USEPA (2014)

US-Avg Albany Region (US SE) Box Elder Region (US NW)

electric ty, coal 45% 52% 31%

electric ty, oil 1% 0% 0%

electric ty, gas 24% 25% 14%

electric ty, nuclear 20% 17% 3%

electric ty, hydro 6% 3% 44%

electric ty, biomass 1% 3% 1%

electric ty, wind 2% 0% 5%

electric ty, solar 0% 0% 0%

electric ty, geothermal 0% 0% 1%

electric ty, other fossil fUel 0% 0% 0%

Note. Predominant fuel sources for each region are bolded

Table A2. Scenarios for Sensitivity Analysis

Scenario Life cycle stage affected Description of Model Change

High forest yield (m3 wood/hec) Forestry stage (for pulp) Increase yield in forestry process by 25%

Low forest yield (m3 wood/hec) Forestry stage (for pulp) Reduce yield in forestry process by 25%

High water loss for pulp Pulp production Increase water lost to evaporation during pulp production by 100%

Low water loss for pulp Pulp production Reduce water lost to evaporation during pulp production by 50%

Use National Average Pulp Energy and Emissions Data Pulp production Modify the pulp production data to use a national average energy and air emissions data in place of existing supplier data

National average electricity LCI Paper towel production Use the Ecoinvent 'electricity, medium voltage, at grid' process in place of the region-specific electricity LCI

Facility-level mass allocation Paper towel production Use a facility-level mass allocation approach in place of the IPSA procedure for determining inputs and output quantities per unit towel

Long distribution distance Distribution Increase distribution distance of paper towel to retailers by 100%

Short distribution distance Distribution Decrease distribution distance of paper towel to retailers by 50%

Alternate material selection for WARM Disposal Use 'Newspaper' as the proxy material for paper towel in the WARM model for estimating disposal stage emissions and energy use

Table A3. Data Quality Scores

Process Data quality Indicators

Source Reliability Completeness Temporal Correlation Geographical Correlation Technical Correlation

Electricity 1 1 1 1 1

Fuel 1 1 1 1 1

Pulp 2 1 1 1 1

Pulp Transport 2 1 1 1 1

Other Facility Purchases 1 1 1 1 1

Distribution 1 2 1 2 4

Disposal 2 1 1 2 4

Table A4. Life Cycle Inventory Data on WARM landfilling model for materials used as a proxy

Amount modeled in WARM per roll (kg) Landfill CH4

WARM Material Biogenic C content (%)a generation (CO2-eq/dry MT) a

Mixed paper (used in baseline scenario) 0.17856 44% 3.18

Newspaper (used in alternative scenario) 0.17856 49% 1.33

Corrugated cardboard 0.0134 49°% 3.48

LDPE 0.0032 0% 0

a (Source: ICF International. 2015)

Table A5. Life Cycle Inventory Data for National Average Pulp - Energy inputs and emissions for 1 metric tonne pulp.

Name Amount Unit

Inputs

electricity, at industrial user 163 kWh

coal 40 kg

residual fuel oil 0.0382 m3

petroleum coke 0.0175 m3

biomass 90 kg

natural gas 159 m3

black liquor - internal recycle 1145 kg

lime mud - internal recycle 204 kg

sludge - internal recycle 0.89 kg

Outputs

National Avg Pulp, at pulp and paper plant 1 tonne

Nitrogen oxides 1.77 kg

Sulfur dioxide 2.03 kg

Carbon dioxide, fossil 726 kg

Carbon dioxide, biogenic 1832 kg

Particulate matter < 2.5ug 0.30 kg

Particulate matter > 2.5ug, <10 ug 0.30 kg

Volatile organic compounds, unspecified 0.20 kg

Hydrogen chloride 0.139 kg

HIGHLIGHTS: Detailed Life Cycle Assessment of Bounty Paper Towel Operations in the United States

This study is a cradle to grave life cycle assessment of Bounty® paper towels produced on selected lines in Albany, Georgia and Box Elder, Utah facilities. The new allocation process called Industrial Process Systems Assessment (IPSA) provides line-specific data for a multi-product, multi-line papermaking plant that is more accurate than facility-level allocation.

Improvements made to US electricity and landfilling data, as well as the use of highly detailed and facility-specific pulp production and papermaking and converting data have made this LCA the most robust and technically accurate study on paper towels published. An endpoint normalization procedure focused the results to fossil fuel depletion, climate change, land use, particulate matter impacts; informing the company how to best concentrate efforts to improve the product's sustainability.

This paper towel LCA is designed as a stand-alone assessment as well as a foundational study generating data that is needed for new integrated sustainability metrics developed by US EPA and P&G.

This study demonstrates the use of new advanced scenario analysis capability using enhancements made to openLCA software.