Comparison of geospatial support in RDF stores: Evaluation for ICOS Carbon Portal metadata

The evolution of World Wide Web (WWW) into semantic web is happening with the aid of standards like Resource Description Framework (RDF), SPARQL and a few others from World Wide Web Consortium (W3C). Over the years, semantic data management technologies have been introduced as software platforms commonly known as RDF stores. Lately these RDF stores have been tested for processing and maintenance of large data sets complying with Linked Data principles. In order to standardize geographic capabilities in these RDF stores, Open Geospatial Consortium (OGC) adopted GeoSPARQL as an extension to SPARQL query. Our study aims to discuss the geospatial capabilities, and the conformance to GeoSPARQL standard, of the five RDF stores: Eclipse RDF4J 2.4.0, Apache Jena 3.9.0, Openlink Virtuoso 7.2.4, Stardog 6.0.1 and GraphDB 8.8.0. Along with the investigation of features, the performance evaluation of these RDF stores has also been conducted by measuring the execution times of a set of GeoSPARQL queries. The evaluation query set consists of non- topological, spatial selection as well as spatial join queries adopted from a spatial benchmark, Geographica. The geospatial component of Integrated Carbon Observation System (ICOS) Carbon Portal (CP) metadata has been used for performance evaluation in order to establish the suitability of the RDF stores for ICOS-CP requirements. Java Programs have been developed in order to interact with all the RDF stores for upload of data and execution of benchmark queries. Some result set disparities amongst the RDF stores as well as variation in performance metrics on different hardware platforms have also been highlighted in our research.The World Wide Web (WWW) is in a state of evolution into a semantic web. In semantic web, the web contents are more meaningful and useable for machines as well as human beings. To achieve this, the World Wide Web Consortium (W3C) has introduced a number of web technology standards including the data model standard, Resource Description Framework (RDF) and the query language standard, SPARQL. Proponents of semantic web have also introduced the Linked Data (LD) principles to promote the online availability of semantic datasets with machine readable interlinks. In order to maintain and utilize the semantic contents as well as LD datasets, software platforms developed over the years are categorized as RDF stores. The Geographic data processing, communication and interoperability require additional capabilities. There is a considerable variation of geospatial features across different products. In the spirit of standardization of these geospatial features across the semantic web data platforms, the Open Geospatial Consortium (OGC) has recommended a spatial standard, GeoSPARQL. In this study we aim to investigate the geospatial capabilities and the consistency of these with GeoSPARQL in the RDF stores. Five platforms have been evaluated in our study, including: Eclipse RDF4J 2.4.0, Apache Jena 3.9.0, Openlink Virtuoso 7.2.4, Stardog 6.0.1 and GraphDB 8.8.0. An important characteristic of data management platforms is the efficiency of query processing. The query performance of the RDF stores has also been assessed in our study. For this purpose a set of benchmark queries has been established and Java programs have been developed to interact with the tested platforms from a programming environment. An important utilization of our study is to assess the suitability of the studied RDF stores for metadata maintenance at Integrated Carbon Observation System (ICOS) Carbon Portal (CP). Results of our study provide an insight into the spatial query processing of the selected RDF stores and also highlight the relevance of these platforms from ICOS perspective