The Airbase is the European air quality dataset maintained by the Environmental European Agency. The dataset is available on the Web, and contains air quality monitoring data for 40 European countries. The multidimensional nature of the data makes it a good t for OLAP (Online Analytical Processing) systems. Moreover, by linking the data to the Semantic Web, we can magnify its value, allowing for more sophisticated data analytics. In this paper, we introduce and describe QBOAirbase, a multidimensional provenance-augmented version of the Airbase dataset. QBOAirbase models air pollution data as an RDF cube, which has been linked to the YAGO and DBpedia knowledge bases.
The Airbase1 is the European air quality dataset maintained by the EEA (Environmental European Agency). The dataset is available on the Web, and contains air quality monitoring data for 40 European countries. The numerical and multidimensional data contained in the Airbase dataset can be e ciently handled by Online Analytical Processing systems (OLAP). Such systems are common in data warehousing or business intelligence scenarios, and are optimized to handle complex aggregation queries on multidimensional data with rare updates. Multidimensional datasets, known as data cubes, consist of a set of observations, e.g., measurements of the concentration of an air pollutant. These observations are described in terms of coordinates in a set of dimensions, e.g., time or location. Observations are the target of OLAP applications.
OLAP applications can bene t considerably from RDF and Linked Data [1{3,
8]. Thanks to the Linked Open Data initiative2, resources from di erent datasets
have been interlinked when they refer to the same real-world concept. Such a
network of datasets constitutes what we call the Semantic Web, and allows us to
see the Web as a giant knowledge base that can be queried, \understood", and
analyzed by software agents. The interest in OLAP on the Semantic Web has
been thrusted by the support for aggregation queries introduced in SPARQL 1.1
{the query language for RDF{, and the publication of the QB vocabulary [
Keeping track of the origin of the data is crucial in a setting with multiple
independent data providers. The provenance of a fact in an RDF data collection
is metadata about the source and the data transformations that led to the
publication of that fact. Such metadata nds application in scenarios such as data
fusion or access control [
We present QBOAirbase, a multidimensional, linked and
provenanceaugmented version of the Airbase dataset in RDF. QBOAirbase represents
air pollution data as a three-dimensional multilevel cube modeled with
QB4OLAP [
QBOAirbase is built upon version 8 of the Airbase dataset1. The original dataset
is a collection of CSV and XML les containing annual concentration
measurements for 238 air pollutants (e.g., SO2, PM10) in 40 European countries from
years 1969 to 2012. Besides the actual measurements, the les also contain
information about the data providers and the monitoring stations. For the latter, this
includes the station's geographic location and technical details about the sensors'
con gurations. The data is accessible via a SPARQL endpoint4. Unlike
QBOAirbase, this dataset is modeled with QB [
Observations. In QBOAirbase an observation maps to a measurement in the original Airbase dataset, that is, the aggregation of a set of measurements for a single air pollutant in an annual time span. QBOAirbase includes measurements for a list of 15 pollutants out of the 238 present in the original dataset. This is the minimal list of pollutants that a country must measure according to EU regulations. The original dataset considers 20 aggregation functions such as the annual mean, the maximum, the 50th percentile, among others. They are all considered in QBOAirbase.
Dimensions. An observation is characterized by its coordinates in the year, station, and sensor dimensions as Figure 1 shows. The edges in the gure de ne the schema properties, i.e., the RDF properties that connect the di erent levels of the cube structure. The station dimension contains three levels: station, city, and country. For some stations we did not have access to the information of the 4 http://semantic.eea.europa.eu/sparql
QBOAirbase: The European Air Quality Database as an RDF Cube
city, hence those stations can only be rolled-up to the country level. We have
manually linked the cities and countries in QBOAirbase to YAGO [
air:schema/inCountry air:schema/station
Station air:schema/inCity
City air:schema/locatedIn Country float air:schema/measure Year air:schema/year
Observation air:schema/sensor Sensor air:schema/component
Component
Each RDF triple in QBOAirbase is augmented with its work ow provenance [
The Airbase's website provides an extensive list of reports in the form of gures, tables, interactive maps, and visualizations. Most of the reports can be generated by OLAP operations on the available data. However, some reports require additional data not present in the dataset. One example is the percentage of urban population resident in areas where pollutant concentrations are higher than the recommend limit values5, which requires the population of the cities and 5 https://www.eea.europa.eu/data-and-maps/figures/urban-population-resident-in-areas-pollutant-limit-target the recommended concentration values of the pollutants. Such information can be obtained from another knowledge base. This justi es our decision of linking QBOAirbase to other sources in the Semantic Web. Moreover, the work ow provenance gives users more control on the data by e.g., restricting reports to data coming from particular institutions.
QBOAirbase is publicly available at http://qweb.cs.aau.dk/qboairbase. We also provide a SPARQL endpoint and a detailed documentation of the cube and provenance design. As Airbase, QBOAirbase is released under the terms of the ODC Open Database License (ODbL).
This research was partially funded by the Danish Council for Independent Research (DFF) under grant agreement No. DFF-4093-00301.