Доклади на Българската академия на науките Comptes rendus de l Académie bulgare des Sciences Tome 65, No 5, 2012 MATHEMATIQUES Informatique APPLICATION OF ONTOLOGIES AND SEMANTIC WEB FOR FACILITATION OF ECOLOGY Alexandre Chikalanov, Stoyan Stoyanov, Mariyana Lyubenova, Velichka Lyubenova (Submitted by Academician V. Sgurev on October 19, 2011) Abstract The application of Ontologies in the description of biology has become widely spread. This paper presents an approach for the application of Ontology in the description of complex ecological categories as Plant Functional Types. The authors illustrate the application of two Ontology approaches of engineering usage of Unified Modelling Language and Semantic Web approach usage of RDF/OWL schema languages. Both approaches are powerful enough to be used as tools for formal description and information dissemination of complex ecological categories. The conversion of both approaches in general is under development in the Object Management Group (OMG). Key words: Ecology knowledge, Plant Functional Type, Ontology, OWL, Unified Modelling Language, UML Introduction. The volume of Ecology knowledge has recently increased in an exponential way. Additionally, in the coming era of Internet of Things, a formal machine readable presentation of knowledge increases significantly and even 599
becomes critically important for building a global platform containing Knowledge, Social and Business information. There are a number of papers which consider Ontology as a general approach in achieving this goal [ 9 11 ]. The aim of this paper is to illustrate an Ontology-oriented approach for Ecology knowledge and a more specific formal presentation of the Plant Functional Type applying engineering approach and RDF/OWL schema languages. The authors have presented an approach for storing complex scientific information in Category-oriented databases [ 12 ]. This paper continues the ideas developed in the cited paper by presenting an approach for the exchange of a complex scientific knowledge in machine readable form. Methods. According to [ 7 ] there are three main approaches in Ontology: Usage of first order logic predicate computing. On the base of a set of symbols, logical operations and arithmetical operations compute whether a statement is true or false. Usage of description logic. Usage of software engineering technique. In the present paper, we propose the application of a software engineering technique by applying UML as a common software engineering language and the respective RDF/OWL languages for translation of the UML presentation in a machine readable form. One of the most important characteristics of Ontology and more specifically RDF and OWL schema, which distinguish them from OO description, is that after defining of classes they can be supplied with additional properties without the necessity to redefine the class itself. In our definitions we will use RDF Schema and OWL language. In this way we may add more and more properties to an already defined class. In our further examples we will define the most essential properties of the presented classes. These properties may be further extended seamlessly. Object. The object of formal description is a plant functional type invented by the authors of this paper. Plan Functional Type (PFT) [ 1 ] can be defined as a set of species with similar reactions against climate changes. The PFT proposed by the authors [ 3 ] is related to stress on growth of trees imposed by adverse climatic factors measured by average annual temperature and precipitations. The underlying methodology for the proposed PFT is the following: Taking samples of tree rings of different tree species from different locations; Computing the average tree rings per tree species per location; Polynomial approximation of the average tree samples; 600 A. Chikalanov, S. Stoyanov, M. Lyubenova et al.
Computing of an index as a ratio between measured and computed width per each year for any sample; Computing of an average growth index per tree species per location; Computing of temperature and precipitation indexes in the same way per location; Marking each year as AWD or ACW or other, for which there is climatic data for each location. A year is marked as AWD when the temperature index is greater than 1 + δ 1 and the precipitation index is less than 1 δ 2. A year is marked as ACW when the temperature index is less than 1 δ 1 and the precipitation index is greater than 1 + δ 2. Fig. 1. A general UML model of a proposed Plant Functional Type Compt. rend. Acad. bulg. Sci., 65, No 5, 2012 601
We will call 1+δ 1 and1 δ 1 lower warm and highest cold temperature thresholds, 1 + δ 2 and 1 δ 2 lower wet precipitation and highest dry precipitation thresholds. One of the possible techniques is similar to the technique of Regression analysis: to identify the relationship between two or more quantitative variables: a dependent variable whose value is to be predicted and an independent or explanatory variable (or variables), about which knowledge is available. Regression analysis is used to understand the statistical dependence of one variable on other variables. The technique can show what proportion of the variance between variables is due to the dependent variable, and what proportion is due to the independent variables. The application of a regression analysis is rooted in an initial credible explanatory model. All the explanatory models are constructed on the basis of a model, such as the following, for linear regression: Y = β 0 + β 1 X 1 + β 2 X 2 + + β k X k + ε, Y f(x, β), where Y is the change that the programme is mainly supposed to produce, X 1 k are independent variables likely to explain the change, β 0 k are constants, and ε is the error term. Reliable data are collected, either from a monitoring system or from an observation system. Partially realized regression techniques can be implemented by an evaluation method together with computing of average growth index per tree species per location. A tree species is considered as AWD sensitive PFT for a specific location when its average indexes have highest correlation with AWD years. The same is true for ACW sensitive PFT. Figure 1 presents a general UML model of a proposed PFT. <owl:class rdf:id= LivingSpecies > </owl:class> <owl:objectproperty:id= SelfOrganization <rdfs:domain rdf:resource= # LivingNature > <owl:objectproperty= Reproduction <rdfs:domain rdf:resource= # LivingSpesies > Fig. 2. Living nature description 602 A. Chikalanov, S. Stoyanov, M. Lyubenova et al.
<owl:class rdf:id= PlantSpecies > <owl:subclassof rdf:resorce = # LivingSpecies /> </rdfs:class> <owl:objectproperty rdf:id= Photosynthesis > <rdfs:domain rdf:resource= #PlantSpecies > <owl:objectproperty rdf:id= SpeciesName > <rdfs:domain rdf:resource= #PlantSpecies /> <rdfs:range rdf:resource= {\&}rdf;literal/> <owl:class rdf:id= TreeSpecies > <rdfs:subclassof rdf:resource= #PlantSpecies > <rdfs:subclassof> <owl:restriction> <owl:onproperty rdf:resource= #SingleWoodenStem /> </owl:restriction> </rdfs:subclassof> </owl:class> <owl:objectprperty rdf:id= SingleWoodenStem > <rdfs:domain rdf:resource= #TreeSpecies /> Fig. 3. Plant Species, Tree Species and Tree Rings presentation We define living nature with two of its properties, namely self organization and reproduction (Fig. 2). Plant species are defined with their most essential property photosynthesis. Each plant species has a unique name. Tree species are a sub class of plant species, which have a single wooden axis. Each tree ring sample is from only one tree species. The objects of class TreeRings consist of tree ring samples. The average values of tree ring indexes for every specific place and for every tree species are contained in class AverageTreeRings. Climatic related data as temperature and precipitation indexes are contained in AverageTemperatures and AveragePrecipitations. The definitions of the last two classes are the same as the definition of AverageTreeRings. Classes AWD and ACD contain adverse warm dry and adverse Compt. rend. Acad. bulg. Sci., 65, No 5, 2012 603
cold wet years respectively. The adverse years are selected according to temperature and precipitation indexes. When the temperature index for a year is higher than a threshold value and the precipitation index is lower than a threshold value, we consider this year as adverse warm dry AWD. We define the class Place as a polygon of geographical coordinates of its vertices. In our ontology each class Place is composed of all tree samples which have been taken, the average tree ring indexes per tree species, the average yearly precipitation and temperature indexes and the sets of adverse years. In nature, places are not related to this data. <owl:class rdf:id= TreeRings > <rdfs:subclassof= DataContainer /> </owl:class> <owl:class rdf:id= AverageTreeRings > <rdfs:subclassof= DataContainer /> </owl:class> <owl:objectproperty rdf:ideateaverageringindexes > <rdfs:domain rdf:resource= #AverageTreeRings /> <owl:objectproperty rdf:ideateaverageringindexes > <rdfs:domain rdf:resource= #AverageTreeRings /> <owl:class rdf:id= AWD > <rdfs:subclassof= DataContainer /> </owl:class> <owl:datatypeproperty rdf:id= LowerTeperatureThreshold > <rdfs:domain rdf:resource= #AWD /> <rdfs:range rdf:resource= {\&}xsd:greaterthan /> </owl: :DataTypeProperty> <owl:datatypeproperty rdf:id= HigherPrecipitationThreshold > <rdfs:domain rdf:resource= #AWD /> <rdfs:range rdf:resource= {\&}xsd:lowerthan /> </owl: :DataTypeProperty> Fig. 4. Definitions of some classes which compose the class Place 604 A. Chikalanov, S. Stoyanov, M. Lyubenova et al.
Ontology( <http://www.w3.org/2001/sw/bestpractices/oep/simplepartwhole/ part.owl> ) ObjectProperty(hasPart_directly inverseof(partof_directly)) ObjectProperty(partOf Transitive inverseof(haspart)) SubPropertyOf(hasPart_directly haspart) SubPropertyOf(partOf_directly partof) Namespace(part = <http://www.w3.org/2001/sw/bestpractices/oep/simplepartwhole/ part.owl#>) Ontology( <http://www.w3.org/2001/sw/bestpractices/oep/simplepartwhole/ example1.owl> Annotation( owl:imports <http://www.w3.org/2001/sw/bestpractices/oep/simplepartwhole/ part.owl>) Class(Place partial restriction(part:haspart allvaluesfrom(unionof(treerings AverageTreeRings AvereageTemperatures AveragePrecipitations AWD ACD)))) Class(TreeRings partial) Class(AverageTreeRings partial) Class(AverageTemperatures partial) Class(AWD partial) Class(ACD partial) ) Fig. 5. Definition of Place and composing classes But bearing in mind the goal of defining plant functional types of tree species on a specific place, this information is inevitable. Without any of the composing objects, a specific place is useless for classifying the trees on its territory. Figure 4 illustrates the definitions of classes which compose the class Place [ 4 ]. Figure 5 presents the definition of class Place. Compt. rend. Acad. bulg. Sci., 65, No 5, 2012 605
<owl:class rdf:id= AdversePlantFunctionalType > <owl:oneof rdf:parsetype= Collection > <owl:thing rdf:about = #AWD_Sensitive /> <owl:thing rdf:about = #ACD _Sensitive /> <owl:thing rdf:about = #Other /> </owl:one of> </ owl:class> Fig. 6. Adverse Plant Functional Type We will define the class TreeRingPlantFunctionalType via direct enumeration of its members. The members AWD Sensitive and ACD Sensitive members denote one of both adverse plant functional types. The member Other denotes that a tree species does not belong to one of these functional types. The evaluation of climatic data and tree rings samples taken from specific places can be done by applying different evaluation approaches like computing stress periods (3) for tree sampled species or by applying regression analyses or any other function for finding the relation between tree rings samples and climatic condition in this place. In Figure 1 we presented a UML diagram which defines a common class for mapping climatic and sample data to one of the introduced Plant Functional Types. In terms of OWL we describe the Evaluation function as a functional property with domain class Place and range class AdversePlantFunctionalType [ 5 ]. Any specific valuation function or algorithm is an instance of this property. Figure 7 contains the description of a functional property which relates the individuals of class Place to the individuals of class AdversePlantFunctionalType, which can be of the enumerated values. Conclusion. As a summary of the proposed usage of Ontologies for formal and machine readable description of Ecology objects, we can make the following conclusions: 1. Application of UML is powerful enough for the description of Ecology categories. <owl:objectproperty rdf:id="hasvintageyear"> <rdf:type rdf:resource="{\&}owl;functionalproperty" /> <rdfs:domain rdf:resource="#place" /> <rdfs:range rdf:resource="#adverseplantfunctionaltype "/> Fig. 7. Functional property mapping Place class to AdversePlantFunctionalType 606 A. Chikalanov, S. Stoyanov, M. Lyubenova et al.
2. In order to describe complex Ecological categories we need to apply OWL DL version of OWL, which is its most powerful version. The conversion between UML and OWL is under consideration and development in the Object Management Group (OMG), which develops an Ontology Definition Metamodel (ODM). In this way, after creating a set of conversion tools, the descriptions will allow us to map them without loss of information. Acknowledgements. The survey are conducted according to the goals of the scientific project with national co-fund entitled Application of information technologies for modelling of forest ecosystems approach to the development of dynamic global vegetation models, COST Action ES0805. REFERENCES [ 1 ] Smith T., H. Shugart, F. Woodward (eds). Plant Functional Types. Cambridge University Press, UK, 1997, 361 p. [ 2 ] Lavorel S., S. Diaz, J. Hans, C. Cornelinessen, Eric Garnier et al. In: Terrestrial Ecosystems in a Changing World, Berlin-Heidelberg-New York, Springer, 2007, 49 164. [ 3 ] Lyubenova M., A. Chikalanov. Compt. rend. Acad. bulg. Sci., 63, 2010, No 3, 409 418. [ 4 ] http://www.w3.org/2001/sw/bestpractices/oep/simplepartwhole/ [ 5 ] http://www.w3.org/tr/owl-guide/#owl_functionalproperty [ 6 ] Lyubenova M. Plant Ecology, Sofia, Marin Drinov Acad. Publ. House, 2004, 574 p. [ 7 ] Gomez-Perez A., M. Fernandez-Lopez, O. Corcho. Ontological Engineering, Springer-Verlag, London Limited, 2004, 403. [ 8 ] OMG Unified Modelling Language (OMG UML), Infrastructure, September 2009. [ 9 ] Klein M., N. F. Noy. A Component-Based Framework for Ontology Evolution. In: Workshop on Ontologies and Distributed Systems at IJCAI-2003, Acapulco, Mexico, 2003. [ 10 ] Todorovski L., S. Dzeroski. Using domain knowledge on population dynamics modelling for equation discovery. In: The 12th European Conf. on Machine Learning, Lecture Notes in Computer Science, Berlin, Springer Verlag, 2001, 478 490. [ 11 ] Fielding J. M., J. Simon, W. Ceusters, B. Smith. Ontological theory for ontological engineering: biomedical systems information integration. In: 9th Int. Conf. on the Principles of Knowledge Representation and Reasoning (KR2004), Whistler, Canada 2004. [ 12 ] Chikalanov A., M. Lyubenova, S. Stoyanov. Compt. rend. Acad. bulg. Sci., 63, 2010, No 9, 1327 1334. Europe for Everyone 126, Stamboliiski Blvd, fl. 7, ap. 12a 1233 Sofia, Bulgaria e-mail: ctmdevelopment@yahoo.com Geophysical Institute Bulgarian Academy of Sciences Acad. G. Bonchev Str., Bl. 3 1113 Sofia, Bulgaria 4 Compt. rend. Acad. bulg. Sci., 65, No 5, 2012 607
Faculty of Biology St. Kliment Ohridski University of Sofia 8, Dragan Tsankov Blvd 1164 Sofia, Bulgaria Technical University of Sofia 8, St. Kliment Ohridski Blvd 1756 Sofia, Bulgaria 608 A. Chikalanov, S. Stoyanov, M. Lyubenova et al.