A Minimal Ontology Pattern for Life Cycle Assessment Data

Abstract

Life Cycle Assessment (LCA) studies the environmental impact of products taking into account their entire life-span and production chain. This requires gathering data from a variety of heterogeneous sources into a Life Cycle Inventory (LCI). LCI preparation involves the integration of observations and engineering models with reference data and literature results from around the world, from di erent domains, and at varying levels of granularity. Existing LCA data formats only address syntactic interoperability, thereby ignoring semantics. This leads to a variety of challenges, e.g., di culties in reproducing assessments published in the literature. In this work, we present an ontology pattern that speci es key aspects of LCA/LCI data models, namely the notions of ows, activities, agents, and products, as well as their properties.

-

Life Cycle Assessment (LCA) is concerned with analyzing the environmental impact of products, taking into account the complete production chain and the entire life-span of the product. For instance, assessing the impacts of operating a solar array goes beyond the pure manufacturing and assembly of the photovoltaic modules. It also includes transportation emissions, installation emissions, operation emissions, and the nal disposal emissions. Such assessment rst requires the gathering of all relevant data from di erent sources into a so-called Life Cycle Inventory (LCI), followed by the actual assessment of the environmental impacts based on the gathered data, used models, and the literature. Understanding the complex impact of products is crucial for arriving at, and maintaining, a sustainable world where human needs are met without causing harm to the environment or impacting the ability of future generations to meet their needs. As such, LCA is a highly interdisciplinary eld that requires synthesizing information from a variety of discipline-speci c studies. This interdisciplinarity can be challenging because the vocabularies often vary between and even within elds of study. This can create signi cant problems for data sharing in LCA when data from multiple sources are translated, merged, and managed as a single life cycle inventory.

Current LCA models do not facilitate semantic interoperability [ 6 ]. While the standardized LCA data formats, e.g., Ecospold and ILCD, do allow for the exchange of data, such formats alone do not guarantee that a dataset from one source can be integrated with another source as there is no consensus on the meaning of central nomenclature for LCA. In other words, the data models are not backed with explicit conceptual models. Ignoring di erences in these underlying assumptions, i.e., settling with syntactic interoperability alone, is likely to cause erroneous and unreproducible results.

Many of the signi cant challenges to data management in LCA practice [ 11 ] and interpretation [ 10, 12 ] arise from the lack of protocols and mechanisms to ensure mutual comparability and consistency of data sets and results. While Linked Data and ontologies hold great promise for addressing such issues, semantic techniques have only recently been introduced [ 2, 9 ] and thus not yet impacted LCA practice. E orts to develop semantically enriched LCA databases [ 1 ] and product models [ 15 ] have had limited scope. Given the high interdisciplinarity and granularity within LCA, arriving at a shared monolithic domain model seems like a distant goal. Thus, in this work we introduce a minimal ontology design pattern[ 3 ] for LCA data to act as a common core. 2

Competency Questions

Developing an ontology requires use cases that capture recurring domain or cross-domain problems. These uses cases can guide the design of the ontology and help in its evaluation. One approach are co-called competency questions [ 4 ]. These are (often informal) queries that the ontology should be able to answer and that act as requirements for its axiomatization. To give a simple example, if a typical subject matter expert would make a distinction between two classes B and C of a common subclass A, then an ontology that does not introduce B would not be considered as suitable. The following listing shows examples of competency questions that have been identi ed by LCA experts. { Is ow x a reference product (e.g., electricity from a power plant)? { How long will a ow or activity persist (e.g., the emission of land ll gas)? { To which compartment does an elementary ow belong to (e.g., soil)? { What is the location of the agent performing the activity in study x (e.g., where is the coal power plant located for which the emissions of electricity production were assessed in the research study)? 3

The Content Ontology Design Pattern

The proposed ontology design pattern1 is meant to form a common core for the semantic description of key elements of life cycle inventories. It does neither cover 1 OWL le at: http://descartes-core.org/ontologies/lca/1.0/LCAPattern.owl the process of carrying out life cycle assessments, e.g., how data is gathered or how system boundaries are de ned, nor does it provide the variety of spatial, temporal, and thematic attributes used to scope inventory items, e.g., to express the fact that coal extraction may have varying impacts depending on the geographic region and used technology [ 14 ]. Instead, our pattern aims at fostering interoperability between existing data models, speci cations, and software, with the intent to act as a joint building block for the rapidly increasing interest in semantics within the broader LCA community. Due to lack of space, we only discuss a few selected axioms here. An overview of the pattern is depicted in Fig. 1.

Estimating the environmental impact of a certain product requires an understanding of all impacts accumulated during its creation, lifetime, and decommissioning. With respect to the solar panel example introduced before, the creation of the solar arrays requires multiple activities such as the transportation of resources, the generation of electric power by a coal power plant necessary to manufacture certain parts of the panels, or the disposal of polluted sludge accumulated during the production. In other words, the Eco-e ciency of solar panels depends on the activities involved in all stages of their life-cycle. Each activity is performed by at least one agent such as an coal power plant that performs the generation of electricity (Eq. 1). An activity is located via the location of the agent performing it (Eq. 2). Activities also have a temporal extent and can have a variety of properties such as the amount of electricity generated, and so forth.

Activity v 8perf orms Agent u 9perf orms Agent

perf orms hasLocation v hasLocation (1) (2)

Flows are streams of material or energy that can act as the inputs and outputs of activities. In our running example, coal is an input to the activity of electric power generation, while CO2 emissions are an output. Typically, emissions are an undesired product of power generation, and, thus, must be distinguished from reference products (Eq. 3) such as the produced electricity, which is also an outcome of the activity. Note that while ows and their activities both have temporal extents, these extents can di er substantially. An activity such as waste

Ref erenceP roduct v P roduct

Besides their role as products, ows can be categorized into elementary ows or intermediate ows (Eq. 4, 5). The rst case describes material or energy that is entering the system from the environment without any previous transformation by humans or is leaving the system by being released into the environment without further human transformation (Eq. 7) [ 7 ]. The environment is typically described in terms of compartments such as air, water, or soil (Eq. 6). In contrast, intermediate ows occur between processes of the studied system. disposal may take hours, while the resulting emissions may continue for years. (3) (4) (5) (6) (7) (8) (9)

IntermediateF low u ElementaryF low v ?

ElementaryF low u Ref erenceP roduct v ?

fair; water; soilg v Compartment

F low u 9hasCompartment:fair; water; soilg v ElementaryF low

Guarded domain and range restriction (see Eq. 8, 9 for examples) enable us to infer additional facts about agents, activities, ows, locations, and so forth.

9isInputOf:Activity v F low

F low v 8isInputOf:Activity

4

Relation to Other Patterns and Ontologies

The pattern can be related to a number of other ontologies and ontology design patterns. Details about properties and their values can be modeled via alignments to the GeoLink property pattern [ 8 ]. The location of an agent can be expressed using GeoSPARQL, e.g., to represent the spatial footprint of a region. The temporal extent can be modeled via OWL:Time. As discussed before, the pattern models activities and ows, e.g., by describing their duration, inputs, outputs, roles, and so forth. Equally relevant for LCA is the temporal and spatial extent that scopes the impact assessments, e.g., to state that a certain level of emissions was representative of Western Europe during the 1980s. This can be modeled using the LCA scope ontology [ 14 ]. Our pattern and the scope ontology can be aligned using their common Flow class. The pattern can further be aligned to the material transformation pattern [ 13 ] that deals with products as outcomes of transformations. Units are handled using the QUDT ontology [ 5 ]. 5

Coal Power Plant Example

Here, we outline an example for emissions from a coal power plant to show how the pattern answers the competency questions. In this case, CO2 emission is a Flow as is the used coal. Electric power generation is modeled as an Activity, the CO2 emission as its output (isOutputOf ), and coal as its input (isInputOf ). The power plant generating (performs ) said electricity is of type Agent and has a certain location. Thereby the activity can be located as well. The CO2 emission is an ElementaryFlow with air as its compartment. In contrast, electric power (which is also an output of the power generation activity) is the ReferenceProduct. The activity and the ows can have a temporal extent (hasTemporalExtent ). In this work, we brie y outlined the motivation behind developing a pattern as common building block for LCA/LCI. We presented a few of the used axioms and gave examples from the domain of power generation. The described work is the joint result of a VoCamp on LCA that brought together leading international domain experts with ontology engineers in March of 2015 at UCSB. Future work will focus on introducing the pattern to a larger audience and integrating it with the ongoing ontology projects at the US Environmental Protection Agency.

References

1. Bertin , B. , Scuturici , V.M. , Risler , E. , Pinon , J.M.: A semantic approach to life cycle assessment applied on energy environmental impact data management . In: Proceedings of the 2012 Joint EDBT/ICDT Workshops . pp. 87 { 94 . ACM ( 2012 )

2. Earthster Core Ontology: http://www.epimorphics/com/web/ projects/ECO

3. Gangemi , A. , Presutti , V. : Ontology design patterns . In: Handbook on ontologies, pp. 221 { 243 . Springer ( 2009 )

4. Gruninger, M. , Fox , M. : Methodology for the Design and Evaluation of Ontologies . In: IJCAI'95, Workshop on Basic Ontological Issues in Knowledge Sharing ( 1995 )

5. Hodgson , R., Keller, P.J.: Qudt-quantities, units, dimensions and data types in owl and xml . Online (September 2011 ) http://www. qudt. org ( 2011 )

6. Ingwersen , W.W. , Hawkins , T.R. , Transue , T.R., Meyer, D.E., Moore , G. , Kahn , E. , Arbuckle , P. , Paulsen , H. , Norris , G.A.: A new data architecture for advancing life cycle assessment . The Intern. Journal of Life Cycle Assessment pp. 1 { 7 ( 2015 )

7. Iso , I.: 14044: environmental managementlife cycle assessmentrequirements and guidelines . International Organization for Standardization ( 2006 )

8. Krisnadhi , A. , Hu , Y. , Janowicz , K. , Hitzler , P. , Arko , R. , Carbotte , S. , Chandler , C. , Cheatham , M., and others: The geolink modular oceanography ontology . In: Proceedings of the 14th International Semantic Web Conference . Springer, LNCS ( 2015 ; forthcoming)

9. Mun~oz, E., Capon-Garc a , E., La nez , J.M. , Espun~a, A. , Puigjaner , L. : Considering environmental assessment in an ontological framework for enterprise sustainability . Journal of cleaner production 47 , 149 { 164 ( 2013 )

10. Plevin , R.J. , Delucchi , M.A. , Creutzig , F. : Using attributional life cycle assessment to estimate climate-change mitigation bene ts misleads policy makers . Journal of Industrial Ecology 18 ( 1 ), 73 { 83 ( 2014 )

11. Reap , J. , Roman , F. , Duncan , S. , Bras , B. : A survey of unresolved problems in life cycle assessment . Intern. Journal of Life Cycle Assessment 13 ( 4 ), 290 { 300 ( 2008 )

12. Suh , S. , Yang , Y. : On the uncanny capabilities of consequential lca . The International Journal of Life Cycle Assessment 19 ( 6 ), 1179 { 1184 ( 2014 )

13. Vardeman , C. , Krisnadhi , A. , Cheatham , M. , Janowicz , K. , Ferguson , H. , Hitzler , P. , Buccellato , A. , Thirunarayan , K. , Berg-Cross , G. , Hahmann , T. : An ontology design pattern for material transformation . In: 5th Workshop on Ontology and Semantic Web Patterns (WOP2014) . pp. 73 { 77 ( 2014 )

14. Yan , B. , Hu , Y. , Kuczenski , B. , Janowicz , K. , Ballatore , A. , Krisnadhi , A.A. , Hitzler , P. , Suh , S. , Ingwersen , W.: An ontology for specifying spatiotemporal scopes in life cycle assessment . In: Diversity++ Workshop at ISWC 2015 ( 2015 )

15. Zhang , Y. , Luo , X. , Buis , J.J. , Sutherland , J.W. : Lca-oriented semantic representation for the product life cycle . Journal of Cleaner Production 86 , 146 { 162 ( 2015 )