Current research evidences an increase of use of Semantic Web technologies within city energy management solutions. Di erent ontologies have been developed in order to improve energy data interoperability. However, these ontologies represent di erent energy domains, with di erent level of detail and using di erent terminology. This heterogeneity leads to an interoperability problem that hinders the full adoption of these ontologies in real scenarios. This paper presents the OEMA (Ontology for Energy Management Applications) ontology network. This ontology is an attempt to unify existing heterogeneous ontologies that represent energy performance and contextual data. The paper describes the OEMA ontology network development process, which has included ontology reuse, ontology engineering and ontology integration activities. The paper also describes the main OEMA ontology network modules.
Energy management in current cities is evolving towards the future Smart Grid. Smart Grid managers aim to improve current grid e ciency, sustainability and resilience through Information and Communication Technologies (ICT)-based Energy Management Systems (EMSs). The e orts to improve energy e ciency concentrate on enhancing EMS to optimize the use of renewable and non-renewable energy sources. Energy sustainability measures include the suggestion of actions to change energy management behavioural patterns for economic, social and ecological purposes. Citizens are the main actors and the interoperability between them and EMS is essential. Regarding resilience, the objective is to avoid and react to power outages caused by power peak periods or natural disasters. Again, human-machine collaboration is crucial.
Current EMSs mainly focus on improving energy e ciency. However, energy sustainability and resilience systems require new data representation and exchange technologies [8]. Sustainability and resilience systems are required to exchange, extract knowledge and make decisions from large volumes of energy data collected at high rates and in most cases in real time. In addition, energy data belong to many domains and includes energy performance data (i.e., energy quantities, energy performance indicators, etc.) and energy-related and contextual data (i.e., buildings/infrastructures data, geographical data, weather data, etc.). ICT-based systems gather and store these data. Traditionally, these systems operate in functional silos and rely on heterogeneous technologies that pose new interoperability challenges for new EMSs. These challenges include the creation of a common energy data representation model and common vocabularies for human-machine interaction.
Current research in energy management shows the use of the Semantic Web to overcome these challenges ([6], [11], [7]). From the beginning of the current decade, Semantic Web technologies have been applied to create ontologies that represent energy data for di erent domains. These ontologies are the knowledge base of energy sustainability and resilience applications. However, not all ontologies represent the same energy data domains and at the same level of detail. In addition, the ontologies use di erent vocabularies to describe the same energy concepts. Hence, there is the need of an uni ed energy ontology that can be used in a wide variety of Smart Grid scenarios (i.e., Smart Homes, microgrids, etc.).
This paper presents the OEMA (Ontology for Energy Management Applications) ontology network. This ontology network is an attempt to unify existing heterogeneous ontologies that represent di erent energy-related data. The OEMA ontology network represents energy performance as well as contextual data. The paper explains the methodology and main steps followed to develop the OEMA ontology network. The paper also explains the OEMA ontology network modularized structure. Finally, the paper discusses learnt lessons from the OEMA ontology network development process and how future EMSs can bene t from the ontology. The paper is organised as follows: Section 2 provides a stateof-the-art of ontologies that represent energy domains. Section 3 emphasises on the need of a standardized energy ontology. Section 4 explains the methodology followed to develop the OEMA ontology network. Section 5 explains the main components of the OEMA ontology network and the energy data it represents. Finally, in Section 6 the conclusions of the research are presented. 2
Within recent research projects and initiatives, semantic ontologies have been proposed to represent energy related data used by di erent Smart Grid EMSs. These systems are deployed in di erent Smart Grid scenarios: Smart Homes, urban environments (i.e., building, district, city, etc.), organizations, microgrids or Virtual Power Plants (VPPs) and Smart Grid Demand Response (DR) management.
The ThinkHome ontology [11] represents home energy consumption,
production and energy contextual data, i.e., building details, weather conditions,
etc. The DEFRAM project ontology [
The ontology EnergyUse 3 represents the following information: home user pro les, home appliances, and Heating, Ventilation and Air Conditioning (HVAC) systems data, home sensors and actuators data, home appliances energy consumption measures and energy tips discussion data. This ontology is the base for a collaborative web platform that is focused on raising home end users' climate change awareness [4]. The ProSGV3 ontology [9] represents the following data: infrastructure data, electrical appliances data, energy generation and storage systems data, weather report data, events, energy production and consumption and information about energy producers and consumers. The nal purpose is to use this ontology as the knowledge base of EMSs focused on improving Smart Grid DR and sustainability by predicting Smart Grid energy consumption and production.
The Mirabel ontology [20] represents di erent energy actors' (i.e., home endusers) energy exibility for speci c devices. The DERI Linked Dataspace [6] represents energy related data from di erent enterprise domains. These domains include the following: enterprise business entities (i.e., employees, products, etc.), enterprise infrastructures energy consumption, energy consumption measurement sensors and business information, i.e., nance, facility management, etc. This linked dataspace is used by an enterprise observatory system that is focused on improving enterprise energy management at di erent levels from both economic and ecological perspectives.
The Km4city ontology 4 represents data domains about cities. These domains include energy, mobility, statistics, street graph, sensors, cultural heritage, etc. Represented energy domains include organisation and weather data. The SEMANCO ontology 5 represents the following energy domains: building energy consumption data as well as associated energy performance indicators (i.e., energy costs), consumed energy sources, building features, building equipment, weather conditions, buildings geographical location, demographic, environmental and socio-economic data. The objective of this ontology is to provide models for urban energy systems to be able to assess the energy performance of an urban area [5]. Finally, the LCC ontology [15] represents di erent buildings energy consumption data. This ontology has been developed to publish energy consumption data about cities' infrastructures as Linked Data. 3 http://www.essepuntato.it/lode/http://socsem.open.ac.uk/ontologies/eu 4 http://www.essepuntato.it/lode/http://www.disit.org/km4city/schema 5 http://semanco-tools.eu/ontology-releases/eu/semanco/ontology/SEMANCO/
SEMANCO.owl?
The energy ontologies reviewed in Section 2 represent di erent energy data domains depending on the Smart Grid scenario where they are applied. Table 1 shows the level of detail with which some of the available reviewed energy ontologies represent the main energy domains. We consider that an ontology represents an energy domain with a high level of detail when it includes a wide variety of terms and complex class hierarchies about that domain. A medium level of detail representation of an energy domain includes fewer terms and less complex class hierarchies. Finally, a low level of detail representation includes only few classes and virtually no class hierarchies.
As we can observe in Table 1, none of these ontologies represents all types of energy performance and contextual data that should be taken into account for energy management and the data are also represented at di erent levels of detail. In addition, di erent terms are used to represent the same energy concepts. ontology ThinkHome SAREF4EE BOnSAI ProSGV3 ontology ontology ontology ontology
[11] [18] [7] [9] energy domain Infrastructure technical data Energy consumption systems data Energy performance data Sensors/actuators data Energy stakeholders' data Weather/climate data Geographical data Environmental data Distributed energy sources data Energy DR operations
H H H H M H M M
L M H M L L M
M L H M L L M L
This term and domain representation diversity, called semantic heterogeneity [12], leads to an interoperability problem that hinders the full adoption of these ontologies in real scenarios. Hence, there is the need of creating a uni ed ontology that represents all energy domains providing a common terminology. This ontology can be a standard knowledge base of EMSs applied in any Smart Grid scenario. Moreover, a uni ed ontology would reduce the e ort spent by energy management developers when creating energy ontologies and enable them to be more focused on application implementation.
An ontology can be developed as a whole or as a set of interconnected ontologies, what is known as an ontology network [19]. An ontology network allows classifying energy data from di erent domains into domain ontologies that can be linked by establishing top-level relations between energy concepts from different domains. Hence, this modularized approach would improve the ontology reusability.
Considering all this, the authors have developed the OEMA ontology network. It represents all identi ed energy domains in di erent existing energy ontologies at a high level of detail. The OEMA ontology network provides a common representation of concepts that belong to di erent energy domains. The following sections explain the ontology development process and the ontology structure. 4
This section explains the OEMA ontology network development process with phases for requirements de nition, ontology selection for reuse, otology implementation and integration and ontology evaluation. The development process has followed the steps and guidelines de ned by [15] and [19]. [15] provides ontology development steps and explains them through an application example that corresponds to an energy ontology. [19] provides the so-called NeOn methodology for developing ontologies and ontology networks. The following subsections explain the di erent OEMA ontology network developing phases. 4.1
Ontology Requirements De nition During this phase, the OEMA ontology network functional and non-functional requirements have been de ned. These requirements have been de ned taking as a reference requirements de ned in [15] and guidelines proposed in [19]. As it is developed to be used by di erent energy management applications, the OEMA ontology network must represent all energy data domains that are shown in Table 1 at a high level of detail. The domains include energy performance data and energy-related contextual data. The next step has been the de nition of the top-level relations among these domains.
The OEMA ontology network non-functional requirements cover the following aspects: { Represented energy domains must be classi ed in di erent sub-ontologies or modules, which are known as domain ontologies. This modularized structure will ease ontology reuse and modi cation when adapting it to di erent energy management scenarios/applications. { One of the goals of the OEMA ontology network is to provide a common representation of energy data. Hence, each ontology element (class, property) must be named using only one term in order to avoid semantic heterogeneity. { The Web Ontology Language (OWL-2 6) and ontology elements naming notation (CamelCase) have also been de ned. 6 http://www.w3.org/TR/2004/REC-owl-guide-20040210/ 4.2
Ontology Selection for Reuse The next step was the selection of existing energy ontologies and terms to be reused during the OEMA ontology network development. The ontologies (include the ones reviewed in Section 2) have been evaluated taking into account the ontology requirements de ned previously. The authors checked whether previous ontologies represent the energy data domains shown in Table 1. The authors also considered the level of detail when representing the energy domains (see Table 1).
The ThinkHome ontology is one of the ontologies that represents most energy domains with a high level of detail. Although it is designed to be used by home energy management applications, the ontology can also be extended to other Smart Grid scenarios, i.e., organisations energy management, microgrids energy management, etc. In addition, the ThinkHome ontology classi es energy concepts in di erent domain ontologies. Due to its completeness and its modular approach, the ThinkHome ontology has been selected to be the base for the OEMA ontology network.
Other energy ontologies have also been selected for reuse in order to
complement OEMA. They have been used to represent missing concepts in speci c
energy domains and to enrich the ThinkHome ontology. Additionally, authors
selected also DBpedia [
These concepts have been selected semi-automatically with the help of the AgreementMaker ontology mapper. AgreementMaker has helped to nd the ontologies that represent the same concepts with di erent terms and class hierarchies. It also has allowed identifying those classes and properties related to a certain term that are represented by one ontology but not by others. Based on the ontology mapper results, unique terms and class hierarchies have been chosen to represent repeated concepts among di erent energy ontologies. The criteria followed to choose terms and class hierarchies has been to select a target class hierarchy that represents each energy concept in the most detail. Terms from other energy ontologies are then selected in order to enrich the selected class hierarchy. This criteria is considered an asymmetric ontology merging method and is used to merge ontologies avoiding overlapping concepts [16]. 4.3
Ontology Implementation The next stage is the implementation/development of the ontology. During the development process, the selected energy ontologies have been merged according to selected concepts and class hierarchies in the previous step. The OEMA ontology network development process has been performed in ve iterations: ontology structure de nition, ontology reuse, adding new information to the ontology, ontology integration and ontology evaluation.
First, the OEMA ontology network base structure was de ned. As detailed in Section 4.2, the ThinkHome ontology is the base of the OEMA ontology network. The ThinkHome domain ontologies have been restructured to adapt them Energy domain Ontology Energy consumption systems data
Reused concepts
HVAC systems, communication appliances, entertainment TohnintkolHoogmye eawnphepirtlgieayngfocaeocsdi,lsio,tiaecsco,eudaseutvitciocmesysastttaieotmne,sd,dedevvoiimcceeost,cioclimgnhmettiawnngodrssky,cstoemmpso,nents,
equipment manufacturer, external and internal equipment SAREF4EE Appliances working modes and power pro les, ontology device manufacturer and device model.
EnergyUse Wearable devices. ontology PornotSoGloVgy3 ccBhloeaadrnygiincngagrdedeedvveiicvceiecsse,asln,igdphretelisenscgitnrsgiycsdatelevamipcspe,slei,anwntaceerttecarainhtemegaeotnirntyg.ddeveviciecse,s, to other Smart Grid scenarios in addition to the Smart Home energy management scenario. This restructuring process consisted of renaming and including new super-classes for domain ontologies. For example, the ThinkHome building ontology represents building physical elements, building internal and external equipment, and building geometrical features. The owl:Infrastructure class has been added to this ontology in order to represent infrastructures (i.e., microgrids) in addition to homes and buildings. Then, all top-level classes of the ThinkHome building ontology that refer to homes and buildings have been included as subclasses of owl:Infrastructure class. As a result, the OEMA infrastructure ontology has been created. This process has been repeated in all of the ThinkHome ontologies and as a result the rst OEMA domain ontologies have been created: OEMA infrastructure ontology, OEMA Smart Grid stakeholders ontology, OEMA external factors ontology, OEMA energy and equipment ontology. Energy concepts represented by each of these ontologies are later explained in Section 5.
After de ning the OEMA ontology network structure, the previously selected energy concepts (see Table 2) and associated statements have been merged and added to each OEMA domain ontology. During the ontology reuse process, the following techniques have been applied: 1. Specialization: adding reused classes and properties of one ontology as subclasses and subproperties of another ontology. For example, infrastructure types reused from other ontologies have been added as subclasses of the OEMA infrastructure ontology owl:Infrastructure class. 2. Generalization: adding reused classes and properties of one ontology as super-classes and super-properties of another ontology. For example, white goods types (i.e., cleaning devices) reused from other ontologies have been added as super-classes of cleaning and cooking white goods of OEMA energy and equipment ontology.
In addition, the following ontology engineering activities have been performed: A Knowledge extension: creating new classes and properties for relating reused concepts with concepts from di erent ontologies. For example, when reusing infrastructure premises (i.e., residential premises, business premises, etc.) in the OEMA infrastructure ontology, the owl:hasPremise property has been created in order to relate the owl:Infrastructure class with reused infrastructure premises classes.
B Renaming ontology elements : renaming reused ontology classes or properties by changing their identi ers. This change has been performed to unify all reused ontology resources naming according to CamelCase notation. C Changing property domains and ranges : adding new domains and ranges to properties of reused ontologies. For example, the owl:contains-Building property from OEMA infrastructure ontology had only the owl:Cam-pus class as a domain. This property has been changed to have also the owl:Infrastructure class as a domain. With this change the ontology asserts that other infrastructures apart from campuses (i.e., power stations) contain buildings. D New ontologies creation: The DBpedia ontology includes many concepts about geographical, persons and organisations data. Adding all these concepts to any of the OEMA ontologies would hinder the ontology maintenance. Hence, two new ontologies have been created: OEMA geographical ontology and OEMA person and organisation ontology. In addition, some units of measure (i.e., volume, currency, etc.) are linked with concepts represented in di erent OEMA ontologies. Hence, a new ontology has also been created in order to modularize this aspect: the OEMA units ontology.
E Knowledge relocation: moving knowledge from one ontology to another. The purpose of this change is to group all concepts of a speci c domain in one ontology in order to improve the OEMA ontology network maintainability. For example, population socio-economic factors have been moved from the OEMA geographical ontology to OEMA external factors ontology. F Ontology Design Patterns (ODP) application : the N-ary ODP [13] has been reused to restructure reused concepts.
The next step was to add missing concepts not represented in the reviewed ontologies that are used in di erent EMSs or are present in well-known standards such as USEF and OpenADR [10]. Some of these concepts are: energy saving tips, Electric Vehicle (EV), energy market roles, etc. This step has led to the creation of a new domain ontology: the OEMA energy saving ontology. This ontology adds infrastructures and equipment energy saving recommendations to OEMA ontology network.
Finally, OEMA ontology network domain ontologies were integrated. The OEMA ontology network .owl le has been created and all domain ontologies have been imported in this le. After importing each ontology, the following tasks have been performed: 1. Ontology linking : the top-level relations between di erent OEMA domain ontologies have been stablished through new properties. 2. Duplication removal : duplicate concepts among domain ontologies have been eliminated. Disjoint relations have been established between classes that use the same terms to describe di erent concepts, i.e., operational device state and country state.AgreementMaker has been used again in order to detect duplicate concetps among domain ontologies. Finally, the OEMA ontology network has been evaluated with the OOPS! Pitfall Scanner [14], which detects common pitfalls made during the ontology development process. According to OOPS! feedback, the OEMA ontologies pitfalls have been corrected. The logical consistency of OEMA ontologies has been also evaluated with the Pellet reasoner [17]. 5
In this section, the OEMA ontology network is described. The OEMA ontology network is made up of eight interconnected domain ontologies. Each ontology represents one or more energy domains. These ontologies are connected by a core ontology. OEMA top-level structure is shown in Figure 1. { OEMA infrastructure ontology 7: contains data about Infrastructures/buildings.
These data include infrastructure/building types (i.e., household, microgrid, etc.), technical data (i.e., material, surface, etc.), spaces data (i.e., oors, rooms, etc.), geometrical data (i.e., oor area), etc. 7 www.purl.org/oema/infrastructure { OEMA energy and equipment ontology 8: represents energy equipment such as building automation system resources (sensors, actuators/controllers and HVAC systems), industrial equipment (i.e., construction and manufacturing equipment), energy generators (i.e., EVs, Home Power Plants, etc.), loads (white and brown goods), power storage/energy carriers, etc. The ontology also represents the energy equipment features such as devices' power curve and power pro le or device state. { OEMA geographical ontology 9: represents geographical data about infrastructures and energy equipment locations. These data includes populated places (i.e., country, city, district, etc.), natural places (i.e., mountain, sea, etc.) and places geographical attributes such as altitude, depth or area. { OEMA external factors ontology 10: captures external factors that can inuence energy usage. These factors include climate type (i.e., alpine, continental, etc.), environmental conditions (i.e., lighting, noise, etc.) household socio-economic factors (i.e., household income, housing price, etc.), people socio-economic factors (i.e., salary, education level, etc.), population socioeconomic factors (i.e., density, main origin, etc.), and weather phenomenon (i.e., temperature), etc. { OEMA person and organisation ontology 11: represents person and organisation data: person and person attributes (i.e., age, gender, etc.), organisation, organisation internal structure (i.e., departments), organisations economic data (i.e., endowment, net income, etc.), person roles in organisations (i.e., role in project, occupation), etc. { OEMA energy saving ontology 12: represents general and personalized energy saving recommendations. { OEMA Smart Grid stakeholders ontology 13: represents Smart Grid stakeholders and roles in the energy market (i.e., energy consumers, energy suppliers, Distribution System Operators (DSOs), etc.) and energy exibility operations, (i.e. market processes, ex-o ers exchange), etc. { OEMA units ontology 14: represents di erent units of measure used by the OEMA domain ontologies. These units of measure include energy units, area units, capacity units, currency, density units, etc. The OEMA units of measurement ontology is reused by OEMA infrastructure, energy and equipment, geographical, and external factors ontologies. 6
This paper presents the OEMA (Ontology for Energy Management Applications) ontology network. This ontology network is an attempt to unify existing 8 www.purl.org/oema/enaeq 9 www.purl.org/oema/geographical 10 www.purl.org/oema/externalfactors 11 www.purl.org/oema/pao 12 www.purl.org/oema/energysaving 13 www.purl.org/oema/sgstakeholders 14 www.purl.org/oema/units heterogeneous energy ontologies that represent energy performance and energy contextual data.
The diversity of terminology and level of detail of energy domains represented by reused energy ontologies have brought the main challenges of the OEMA ontology network development process. Hence, reused ontologies concepts selection and ontology engineering activities for linking concepts of di erent energy ontologies have been the most complex and time-consuming tasks for the ontology network development process. Energy consumption systems data and infrastructure data are energy domains that present more heterogeneity among existing energy ontologies. Thus, the energy concepts selection task has been particularly challenging when selecting terms that belong to these energy domains. Most concepts and statements from existing energy ontologies have been copied into the OEMA ontology network instead of importing them. Hence, changes that reused ontologies may su er in the future will not have an impact on the OEMA ontology network. However, new concepts added to reused ontologies must be analysed in order to include them or not in the OEMA ontology network.
The OEMA ontology network is made up of eight interconnected domain ontologies. Each ontology represents one or more energy domains. These domains include both energy performance and contextual data. The OEMA ontology network will enable energy applications to extract knowledge and make decisions about large volumes of energy data from di erent domains. In addition, the modularized approach of the OEMA ontology network approach will facilitate reuse and modi cation when adapting it to di erent Smart Grid scenarios and energy management applications. The main contributions of the OEMA ontology network are: 1. Common representation of energy domains: it represents in a uni ed manner and at a high level of detail energy domains that are described by existing energy ontologies using di erent vocabularies and varying levels of detail. 2. Integration of di erent energy domains: it represents all energy domains captured in di erent energy ontologies in a single ontology network.
Considering these bene ts, the OEMA ontology network provides a starting point for a widely accepted energy ontology. Future work will focus on the OEMA ontology network validation by implementing it in real Smart Grid scenarios, i.e., microgrid energy management, home energy management, etc. 3. Brickley, D., Miller, L.: Foaf vocabulary speci cation 0.91 (2007) 4. Burel, G., Piccolo, L.S., Alani, H.: Energyuse-a collective semantic platform for monitoring and discussing energy consumption. In: International Semantic Web Conference. pp. 257{272. Springer (2016) 5. Corrado, V., Ballarini, I., Madrazo, L., Nemirovskij, G.: Data structuring for the ontological modelling of urban energy systems: The experience of the semanco project. Sustainable Cities and Society 14, 223{235 (2015) 6. Curry, E., Hasan, S., O'Riain, S.: Enterprise energy management using a linked dataspace for energy intelligence. In: Sustainable Internet and ICT for Sustainability (SustainIT), 2012. pp. 1{6. IEEE (2012) 7. Daniele, L., Solanki, M., den Hartog, F., Roes, J.: Interoperability for smart appliances in the iot world. In: International Semantic Web Conference. pp. 21{29.
Springer (2016) 8. Finger, M., Portmann, E.: What are cognitive cities? In: Towards Cognitive Cities, pp. 1{11. Springer (2016) 9. Gillani, S., Laforest, F., Picard, G.: A generic ontology for prosumer-oriented smart grid. In: EDBT/ICDT Workshops. pp. 134{139 (2014) 10. Hippolyte, J., Howell, S., Yuce, B., Mourshed, M., Sleiman, H., Vinyals, M., Vanhee, L.: Ontology-based demand-side exibility management in smart grids using a multi-agent system. In: Smart Cities Conference (ISC2), 2016 IEEE International. pp. 1{7. IEEE (2016) 11. Ko er, M.J., Reinisch, C., Kastner, W.: A semantic representation of energyrelated information in future smart homes. Energy and Buildings 47, 169{179 (2012) 12. Maree, M., Belkhatir, M.: Addressing semantic heterogeneity through multiple knowledge base assisted merging of domain-speci c ontologies. Knowledge-Based Systems 73, 199{211 (2015) 13. Noy, N., Rector, A., Hayes, P., Welty, C.: De ning n-ary relations on the semantic web. W3C working group note 12(4) (2006) 14. Poveda-Villalon, M., Gomez-Perez, A., Suarez-Figueroa, M.C.: Oops!(ontology pitfall scanner!): An on-line tool for ontology evaluation. International Journal on Semantic Web and Information Systems (IJSWIS) 10(2), 7{34 (2014) 15. Radulovic, F., Poveda-Villalon, M., Vila-Suero, D., Rodr guez-Doncel, V., Garc aCastro, R., Gomez-Perez, A.: Guidelines for linked data generation and publication: An example in building energy consumption. Automation in Construction 57, 178{ 187 (2015) 16. Raunich, S., Rahm, E.: Towards a benchmark for ontology merging. In: OTM
Workshops. vol. 7567, pp. 124{133 (2012) 17. Sirin, E., Parsia, B., Grau, B.C., Kalyanpur, A., Katz, Y.: Pellet: A practical owldl reasoner. Web Semantics: science, services and agents on the World Wide Web 5(2), 51{53 (2007) 18. Stavropoulos, T.G., Vrakas, D., Vlachava, D., Bassiliades, N.: Bonsai: a smart building ontology for ambient intelligence. In: Proceedings of the 2nd international conference on web intelligence, mining and semantics. p. 30. ACM (2012) 19. Suarez-Figueroa, M.C.: NeOn Methodology for building ontology networks: specication, scheduling and reuse. Ph.D. thesis, Informatica (2010) 20. Verhoosel, J., Rothengatter, D., Rumph, F., Konsman, M.: An ontology for modeling exibility in smart grid energy management. In: eWork and eBusiness in Architecture, Engineering and Construction-European Conference on Product and Process Modelling, ECPPM 2012, 25-27 July 2012, Reykjavik, Iceland, 931-938 (2012)