Sie sind auf Seite 1von 6

More Emerging Issues for Life Cycle Assessment Data and

Databases
Joyce Cooper University of Washington, cooperjs@u.washington.edu
Abstract. Continuing development of data parameterization algorithms for the US
Department of Agriculture LCA Digital Commons has investigated the usefulness and
completeness of imbedded models. An example uses data and algorithms from the US
Environmental Protection Agency NONROAD model which estimates fuel use and air pollution
inventories for agricultural off-road equipment. Three steps were used to prepare the unit
process datasets. First, the NONROAD model was parameterized. Second, data and
algorithms were added to fill data gaps. Third, data representing individual equipment
instances were aggregated to form complete datasets. Although the resulting datasets,
formatted in ecospold v1-v2 and ILCD, are intended to be ready for use in LCA software,
extensive use of parameterization as well as improved meta data (e.g., data quality and
uncertainty information) simultaneously present compatibility issues and opportunities for
additional capabilities.
Introduction. The LCA Digital Commons (see http://www.lcacommons.gov/) is an open
access database being built by the United States Department of Agriculture (USDA) National
Agricultural Library in response to a national need for data representing US agricultural
products for use in Life Cycle Assessment (LCA). These data are to support policy
assessment, technology implementation decision-making, and publically disclosed comparative
product or technology assertions. The Commons database currently contains 466 single-year
and 70 multi-year datasets representing the production of corn, cotton, oats, peanuts, rice,
soybeans, and wheat at the state level. The database will ultimately include data representing
a wide range of bio-products developed by researchers throughout the US and representing all
life cycle stages. In the 2012 ISSST paper Data uncertainty and quality: emerging issues for
Life Cycle Assessment data and databases (Cooper, 2012), data development for the
Commons and emerging LCA database efforts in general were described within the context of
improving meta data that describe uncertainty and data quality. Here an update is provided,
with a focus on data development and software integration efforts.
Research Goals. The goal of this work was to develop unit process data for the Commons by
integrating an existing computational model within an LCA software compatible dataset (i.e.,
formatted in ecospold v1-v2 and ILCD and thus ready for use in LCA software). Thus the intent
was to parameterize the LCA datasets, with parameterization referring to the practice of
including raw data and formulas instead of computed results in each dataset. This means that
data users can view all raw data and their associated uncertainty distributions as well as all
calculations using the raw data when they view the dataset. As described by Cooper, Noon,
Proceedings of the International Symposium on Sustainable Systems and Technologies (ISSN 2329-9169) is
published annually by the Sustainable Conoscente Network. Melissa Bilec and Jun-ki Choi, co-editors.
ISSSTNetwork@gmail.com.
Copyright 2013 by Joyce Cooper. Licensed under CC-BY 3.0.
Cite As:
More Emerging Issues for Life Cycle Assessment Data and Databases. Proc. ISSST, Joyce Cooper. http://
dx.doi.org/10.6084/m9.figshare.810440. v1 (2013)
Copyright 2013 by Joyce Cooper, University of Washington
and Kahn (2012), the benefits of data parameterization are transparency (the raw data and
computations can be clearly documented and reviewed), enhancement of the potential to
represent process variants (e.g., variations in load, process efficiency, etc. can be represented),
and enhancement of interpretation capabilities (e.g., sensitivity analysis can be performed to the
level of internal variables; results can be interpreted as a function of time; global parameters
e.g., climate data can link inventory datasets to each other and to impact assessments).
Investigative Method. Data and algorithms from the US Environmental Protection Agency
(EPA) NONROAD model were used as the foundation for the development datasets
representing the operation of tractors, mowers, balers, combines, dumper/ tenders, generator
sets, irrigation sets, pumps, skid steer loaders, sprayers, swathers, tillers, and
tractor/loader/backhoes. NONROAD algorithms estimate fuel use and air emissions for a wide
range of past, present, and future equipment (see http://www.epa.gov/oms/nonrdmdl.htm) and
are organized using 10-digit source category codes (SCCs) to represent combinations of
exhaust and evaporation technology types for a given type of equipment, fuel, and power range.
Three steps were used to prepare unit process datasets. First, the NONROAD data and
algorithms were parameterized. Second, data and algorithms from other sources were added to
fill data gaps. Third, data representing individual equipment instances (with each instance
representing a specific power/ SCC/ operating year/model year combination) were aggregated
to form complete datasets.
Results.
Parameterization. Data and algorithms from 15 documents (USEPA 2004a-c, 2005a-e, and
2010a-g; available at http://www.epa.gov/otaq/nonrdmdl.htm#techrept) were reviewed. The
review revealed a need for two modifications, intended to improve upon the underlying
NONROAD data and algorithms:
1. Modification of the fuel carbon mass fraction. The NONROAD model represents
equipment using gasoline, diesel, compressed natural gas (CNG), and liquefied
petroleum gas (LPG) and assumes that the fuel carbon mass fraction is 87% for all fuels.
In the preparation of the datasets, the fuel carbon mass fraction was modified to reflect
the carbon fractions used in the Argonne National Laboratorys GREET model (Argonne
National Laboratory, 2012). Specifically, values of 86.3%, 86.5%, 82%, and 72.4% are
used for gasoline, diesel, LPG, and CNG respectively.
2. Modification of fuel consumption estimates. The NONROAD model estimates
carbon dioxide (CO
2
) emissions based on the balance of the carbon in the fuel entering
the engine less only that leaving as total hydrocarbon (THC) emissions but omitting
consideration of the carbon in carbon monoxide (CO) emissions. This carbon
imbalance in NONROAD is most apparent for smaller gasoline engines. Further
inspection of select smaller engine technology types also reveals that the CO emission
factors are 2-3 orders of magnitude larger than others (and interestingly up to 4 orders
of magnitude larger for NOx emission factors). These same engines however have
similar or even lower break specific fuel consumption (BSCF) factors. The high
emission factors are indicative of incomplete combustion and should instead be
associated with higher fuel consumption for the same power output. Given these
findings, in the preparation of all datasets, the remedy chosen to address both the
carbon imbalance and the lower than expected fuel consumption was to keep the
NONROAD emission factors intact and to increase the NONROAD fuel consumption
estimates to include the carbon in the CO emissions.

Copyright 2013 by Joyce Cooper, University of Washington
Next, the NONROAD data and algorithms with the two modifications were entered into MS
Excel as depicted in Figure 1, parameterized as an individual equipment instance using lookup
functions to find the fuel use and emission factors for the group of exhaust technologies
matching the equipment instance/SCC of interest. Raw data (e.g., BSFC and emission factors)
are explicitly listed and used within equations/ mathematical expressions developed to be
readable by OpenLCA (as described at
http://openlca.org/documentation/index.php/Parameter_functions_in_openLCA). Exchanges
(used in the LCA software between unit processes or to/from the environment) then refer to the
parameters representing the final flow quantities to be used in the life cycle inventory.



Figure 1: Example parameterization of nonroad equipment data. Raw data are used in equations within the
dataset. Exchanges then refer to parameters representing the final flow quantities to be used in the life cycle
inventory.


Filling data gaps. Table 1 presents the list of information needs used to identify data gaps
beyond the information provided in NONROAD. Because NONROAD estimates only fuel
consumption and select air emissions (exhaust and evaporative hydrocarbons (HC), CO, oxides
of nitrogen (NOx), particulate matter (PM; as PM10 and PM2.5), sulfur dioxide (SO
2
), and CO
2
),
there was a need to prepare additional parameters. Thus, data and algorithms from a range of
sources were used to estimate the average, minimum, and maximum ambient operating
temperatures; to divide hydrocarbon emissions among methane (CH
4
), ethane (C
2
H
6
),
aldehydes, and other hydrocarbons; to estimate nitrous oxide (N
2
O) air emissions from fuel use;
to add lubricant use, air emissions, and management; and to add equipment production and
storage, spare parts production, and the treatment of used equipment (i.e., equipment
retirement). Further, because NONROAD estimates fuel use and air emissions without
consideration of uncertainty, a range of sources were used to represent variability in equipment

Copyright 2013 by Joyce Cooper, University of Washington
operating temperatures; fuel carbon mass fraction, fuel consumption, and CO
2
emissions;
equipment load factors; fuel sulfur and ethanol contents; and air emissions.

Aggregation of individual equipment instances to form complete datasets. Finally, given the
capability to model hundreds of thousands of types/ model years of equipment, two types of
datasets were prepared. First, vintage year datasets provided data representing a specific type
of equipment for a specific power range and operating in the year in which it is produced (e.g., a
75-100 HP diesel agricultural tractor produced in 2013 and operating in 2013). Vintage year
datasets were developed with the intent of demonstrating a single instance of the model.
Second, fleet datasets provided data representing the mix of equipment operated in a given
year as defined by fuels, power ranges, and model years used in a given operating year (i.e.,
operating in 2012, a fraction of the equipment used was produced in 2012, 2011, 2010, etc.)
based on the NONROAD scrappage curves. These datasets represented a minimum specified
power to account for the use of higher but not lower powered equipment to provide the work
needed for a particular activity (e.g., whereas a 75 HP baler can be used when a 50 HP baler is
needed, a 40 HP baler cannot) and operation in temperature bins (to represent the average,
minimum, and maximum ambient temperatures of operation irrespective of location within the
crop production regions of the US). Fleet datasets were developed with the intent of being
compatible with the crop production data already in the LCA Digital Commons and for use by
practitioners investigating a wide range of activities.


Table 1. Equipment operation information needs
Products and
co-products
Work performed Co-production
Flows from the
environment
Water withdrawal Air (e.g., used in combustion)
Other directly extracted resources (as
opposed to purchased resources; e.g., for
extraction processes)
Technosphere/
intermediate
flows
Energy use
Product materials use (feedstocks that
are incorporated into products and co-
products) and waste management
Ancillary materials use (lubricants,
catalysts, etc.) and waste management
Transport/ distribution of product and
ancillary materials and energy
Equipment construction
Equipment facility use (for storage or
otherwise)
Equipment spare parts
Equipment retirement
Flows to the
environment
Operating emissions including
unrecovered product and ancillary
materials




Discussion. The data described here will be made available through the USDA LCA Digital
Commons in ecospold versions 1 and 2 and the International Reference Life Cycle Data System
(ILCD) data formats. All three are data exchange formats based on XML (eXtended Markup
Language) and related technologies (XSL, XSLT, Schema), are ISO/TS 14048 compliant, and
are usable by a range of LCA software tools and databases.


Copyright 2013 by Joyce Cooper, University of Washington
The extensive use of parameterization and the inclusion of uncertainty information resulted in
data format/ software compatibility issues. First, whereas the ecospold v1 format does not
support parameterization, the ecospold v2 and ILCD formats do. Thus, all parameterized raw
data and equations can be contained within the datasets in the ecospold v2 and ILCD formats
but not in the ecospold v1 formatted data. Further, because the parameterized raw data and
equations have been developed to be readable by OpenLCA slight modifications are required to
ensure they are readable by other LCA tools. For example, among the constants, operators,
and functions recognized by OpenLCA, some are not recognized by GaBi. Specifically, the
constants pi (the ratio of the circumference to the diameter of a circle) and e (base of the natural
system of logarithms) are not supported in GaBi but can be defined as global parameters in a
given GaBi database. Also, the operators div (integer division); mod (modulus); <> (not equal
to); & (logical and); and | (logical or) and the functions and(x1;x2;...xN) (logical and);
avg(x1;;xN) and mean(x1;;xN) (the arithmetic mean of the given numbers); frac(x) (the
fractional part of x); ipower(x;y), power(x;y), and pow(x;y) (raises x to a power of y); not(b) (the
logical complement of b); or(x1;x2;...;xn) (logical or); random() (returns a random number
between 0 and 1); and trunc(x) (the integer part of x) are not supported in GaBi. This is not
always a problem, for example the power functions can very often be handled using the ^
operator and the trunc(x) function can very often be handled using int(x) in both OpenLCA and
GaBi but the ceiling and base 10 logarithm functions are represented differently. Further, there
are a number of functions recognized by GaBi that are not recognized by OpenLCA, such as a
range of economic functions, such as those used to model depreciation.

For data uncertainty, the ideal situation allows uncertain parameters (i.e., those presented as
raw data such as fuel use or emission factors) to be recognized in LCA software as such. This
would yield a wealth of opportunity for sensitivity analysis at the parameter level. This promises
to be very powerful especially if parameters are globalized from inventory analysis through
impact assessment (e.g., if environmental conditions such as ambient temperatures and
precipitation are linked throughout an LCA). However, some software (e.g., OpenLCA) currently
only recognize uncertain exchanges as opposed to uncertain parameters used to estimate
exchanges. This means that linkages to parameters are lost in any sensitivity or uncertainty
analyses.

In parallel with on-going parameter formatting efforts to manage these issues, documentation of
the nonroad equipment data preparation methods were peer reviewed during the winter of 2013.
Among the comments received, the reviewers requested greater disaggregation of the
combustion emissions than had been provided. For this, it was noted that the USEPA
calculates aldehydes and other toxic pollutant emissions from nonroad equipment by using the
National Mobile Inventory Model (NMIM, at http://www.epa.gov/otaq/nmim.htm) as a front-end to
the NONROAD Model. Within this context, NMIM uses ratios of aldehydes and other toxics to
volatile organic compounds which will be integrated into the datasets.

Also, the peer reviewers were asked to comment on the aggregation of individual equipment
instances to form complete datasets and to suggest alternative methods. They noted that for
individuals considering purchasing a new piece of equipment, an average of the full life of the
equipment might be desirable, perhaps with a dependency on how well the equipment is
assumed to be maintained, etc. For supply chain or regional/national-level assessments (e.g.,
for use by industry associations, agro-feedstock users, or policy analysts), it seems some
representation of an on-going fleet (e.g., fleet data over 20+ year time frame) would be of
interest. Presumably these could also be used with data representing the full life of a new piece
of equipment, as part of an improvement analysis.


Copyright 2013 by Joyce Cooper, University of Washington
Further, as part of the peer review it was noted that the scrappage curves in the NONROAD
Model are global averages intended for broad representation of the fleet, at the county or larger
region. Since a particular farm or farm corporation would have a unique set of model years for
each type of equipment, it is likely that this mix would not be well represented by the scrappage
curves in the NONROAD Model nor by any generalized scrappage curves. Thus, for
practitioners attempting to characterize a private fleet, it would be important to have emission
factors for all model years, including very old ones.

In retrospect, the method used to prepare the agricultural equipment data provided a framework
for the adaptation of a wide range of computational models for preparation of both inventory and
impact characterization datasets. Irrespective of the data/model chosen as a basis, data
development can proceed from a review/ update of the raw data and algorithms to
parameterization keyed to inter-software compatibility, the identification and filling of data gaps,
and final aggregation considering a range of possible data uses. For the nonroad equipment
data described here and the USDA Digital Commons data under development at the University
of Washington Design for Environment Lab (representing irrigation and manure systems,
synthetic fertilizer and pesticide production) engagement of the software developers (OpenLCA,
GaBi, and SimaPro) is ongoing.

Acknowledgements. This work was funded by a grant from the United States Department of
Agriculture National Agricultural Library (cooperative agreement Univ_Washington_SCA 6-29-
10). The author thanks NALs Dr. Simon Liu, Susan McCarthy, and Gary Moore for their
support and guidance and Sarah Boyd of PE International for providing information used in the
comparison of OpenLCA and GaBi functions.

References
Argonne National Laboratory (2012) The Greenhouse Gases, Regulated Emissions, and Energy
Use in Transportation Model, GREET 1 2012 rev1, accessed July 2012 at
http://greet.es.anl.gov/
Cooper J.S. 2012. Data uncertainty and quality: emerging issues for Life Cycle Assessment
data and databases. Proceedings of the 2012 IEEE International Symposium on Sustainable
Systems and Technology, May 16-18, Boston, MA.
Cooper J.S., Kahn E. 2012. Commentary on issues in data quality analysis in life cycle
assessment. International Journal of Life Cycle Assessment, Issue: 17(4) pp: 499-503.
Cooper J.S., Noon M., Kahn E. 2012. Parameterization in Life Cycle Assessment Inventory
Data: review of current use and the representation of uncertainty. International Journal of
Life Cycle Assessment, Issue: 17(6) pp: 689-695.
Cooper J.S., Kahn E., Ebel R. 2013. Sampling error in U.S. field crop unit process data for Life
Cycle Assessment. International Journal of Life Cycle Assessment, Issue: 18(1) pp: 185-
192.

Das könnte Ihnen auch gefallen