This work presents the BLO ontology (Batata Lake Ontology), an application ontology that describes in a structured way the data of research done by limnology researchers of Federal University of Rio de Janeiro (UFRJ) Macaé in Batata Lake (Oriximiná/PA). The main contribution of the BLO is the creation of a research data repository in RDF and the BLS application (Batata Lake System), a semantic web application to support researchers in environmental impact assessments, in preservation areas settings, in species protection and recovery of degraded areas, among other activities.
implementing systems, the practical part [
(www.daml.org/ontologies/), Protégé_Ontology_Library (protegewiki.stanford.edu/wiki/Protege_Ontology_Library), Schemapedia (datahub.io/pt_BR/dataset/schemapedia) and Swoogle (swoogle.umbc.edu/), in order to find ontologies related to this work. The ontologies HydroBodyOfWater (sweet.jpl.nasa.gov/2.0/hydroBodyOfWater.owl) and
features related to the proposed ontology, but they do not address the domain of this
work. After the BLO definition,
Figure 1 shows the graph preview of the main classes of the BLO. The vertices are classes or concepts defined in the ontology. The edges, which have a one direction, are the relations between classes, also called object properties. The Sampling class describes the collected sample by the researcher in the sampling stations, represented by SamplingStation class. SamplingStation has two data properties: coordenates and impacted, which respectively specify the geographical location of the sampling station and whether it is an impacted area or not. The object property isDoneOn determines the relation between Sampling and SamplingStation. The relation isDoneDuring between
number of possible relations is limited by the amount of sampling stations that had some collected sample. The FloodPulse class specifies the lake flood pulses, which are the process stages of filling and emptying of the lake. This class has no data property, because the identification of instances is done by the URI (Flood, HighWater, ebby,
and year in which the sampling is done. It is related to FloodPulse class by the object property determines. This property describes the relation between the months of the year and the flood stages of the lake, which can suffer changes over the years, because there is no standard in the establishment that a month will have a particular flood pulse.
SuspendedMatterial, Aluminum, Chorophyll, Iron, Nitrongen, Oxigen and Phosphor.
The object property isDoneOn between Sampling and SamplinStation is defined with the restriction FunctionalProperty. Thus, a sampling x can be done in only one sampling station y. Using the triple Sampling-> isDoneOn-> SamplingStation is possible seek sampling information grouped by sampling stations. The object property determines is defined as inverse of isDeterminedBy. It allows that when answered the competency question "What is the flood pulse of a certain period?", the reasoner identifies the inverse relation isDeterminedBy and retrieve any instance that has the inverse as relation. Restrictions like these add semantic details to the data model and with reasoners the queries can obtain more accurate results, as shown in the section 3.
stored in the last 26 years in spreadsheets. These data were automatically exported to
3.
for researcher analysis. It was implemented in JAVA with JENA library (http://www.w3.org/2001/sw/wiki/Jena), which allows connecting the application to the
has support for manipulating RDF triples, OWL, SPARQL [
done in a period or by a particular researcher in impacted area or not. It answers the competency question "Which samplings were done in impacted areas in a given period?". The application can consider the filter by researcher, otherwise it will be considered by the period. The samplings can be selected by sampling stations. The
station type that the researcher wants to get as answer in the sampling query page. The query result helps to evaluate the samplings which were done in impacted areas and thus comparing with samples done in non-impacted areas, in order to historically evaluate the behavior and recovery of the environment.
In order to evaluate the data model defined by BLO and the BLS application, a small exploratory study was conducted. The hypothesis was that the use of Semantic Web technologies for describing the Batata Lake data would facilitate the access and analysis of these data. The study was performed from a test divided into two stages: the execution of a search activity using the BLS application and the fill of an evaluation questionnaire. The activity was evaluating the water turbidity of a sample in a given period, considering as parameter the sampling data of non-impacted areas done in the same period. This is important for the researchers, since that allows evaluating the progress of the lake recovery. The study involved seven participants. The choice of them was premised on the experience and engagement with lake researches. Two of the participants, one PhD researcher and one master student, accompanied and provided all the necessary for understanding the domain and definition of competency questions.
The goal of the study was to evaluate how the research data started to be searched and analyzed using the BLS. It was not stipulated time for performing the activity. At the end of the activity each participant filled a joint questionnaire with the following questions: 1) Do the searches available in the web application allow finding and relating the data of the samplings? Why? 2) Do the results obtained by the searches facilitate the comparison of the data and the analysis of the lake recovery? Why? 3) Would you use this application again to query and analyze your research data? Why? 4) Do the terms and system’s menu options correspond to the everyday reality of research about the lake? If the answer is no, list the terms that do not match the reality. 5) Considering a scale of one to five, with option 1 equal bad and 5 equal great, how do you rate the form of searching available in the web application, comparing it with that currently performed in Excel spreadsheets?
Most participants (five of them) answered "yes" to the questions and valued the new way to query research data. Six participants said that would use the BLS application again, as this tool significantly reduces the time spent looking for a data, enabling faster analysis. Six of them said that the vocabulary was defined according to the everyday reality of research about the lake. This indicates that the BLO ontology was well defined according to the domain. The test results also allowed identifying problems and difficulties in finding and analyzing the data. In the issue 2, the answers of four participants indicated that the queries results did not facilitate the data comparison and the lake recovery analysis, because the way the results were presented. They informed the search filter by period should be only for year interval with a flood pulse filter to facilitate analysis based on different periods and years. In addition, they suggested the choice of some variables, such as turbidity or chlorophyll, presented in parallel all the values separated by sampling stations, impacted or not. It would allow analyzing a historical series of data and effectively evaluate the lake recovery. Plus, they observed that the application navigability would be more intuitive with the access to the samplings from the data of a given period.
This paper presented the BLO ontology for semantically describing data of the research done by limnology researchers of UFRJ-Macaé on Batata Lake. The semantic description of these data enables richer queries about the lake through inferences done by reasoners. In addition, it provides a vocabulary of common terms used in other researches about the Batata Lake. The main contributions of this ontology is the creation of a research data repository in RDF and the development of the BLS system, a semantic web application to support researchers to query and analysis the research data about this lake. The aim is supporting the production of scientific knowledge from the analysis made by semantic queries and preparing the data for online publication when needed. An initial exploratory study was done to validate the ontology and the application. The tests showed the BLO relevance and quality and some necessary changes in the BLS application. After implementing these changes, a new experiment will be conducted to validate them.
W3C.