Publishing Skopje Air Quality Data as Linked Data

Abstract

Publishing raw data as Linked Open Data gives an opportunity of data reusability and data understandability for the computer machines. Today, the air pollution problem is one of the biggest in the whole world. Republic of Macedonia, especially its capital Skopje, has big problems with the PM2.5 and PM10 particles in the air approved by several measurement stations positioned on several locations in Skopje. In this paper, we demonstrate the process of centralizing of all the data collected from different measurement stations in one database. Also, we enable interpolation of collected data providing information about the current air quality state in the area between the measurement stations using previously implemented eco models. Interpolated data is saved in the same database providing interfaces that transform saved data into four-star and five-star data, by reusing the existing ontologies from the domain and linking them to the physical places where the measurements were taken and the interpolations were calculated. As a use case scenario, we provide and heat map about the values from various pollutants in the areas in Skopje providing information about the regions that have problems with air pollution.

Full text

Publishing Skopje Air Quality Data as Linked Data
Kostadin Mishev1, Angjel Kjosevski1, Nikola Kalemdzhievski1, Nikola Koteli1, Milos Jovanovik1,
Kosta Mitreski1, Dimitar Trajanov1
1Faculty of Computer Science and Engineering
Ss. Cyril and Methodius University
Skopje, Macedonia
Abstract— Publishing raw data as Linked Open Data gives an
opportunity of data reusability and data understandability for
the computer machines. Today, the air pollution problem is one
of the biggest in the whole world. Republic of Macedonia,
especially its capital Skopje, has big problems with the PM2.5
and PM10 particles in the air approved by several measurement
stations positioned on several locations in Skopje. In this paper,
we demonstrate the process of centralizing of all the data
collected from different measurement stations in one database.
Also, we enable interpolation of collected data providing
information about the current air quality state in the area
between the measurement stations using previously implemented
eco models. Interpolated data is saved in the same database
providing interfaces that transform saved data into four-star and
five-star data, by reusing the existing ontologies from the domain
and linking them to the physical places where the measurements
were taken and the interpolations were calculated. As a use case
scenario, we provide and heat map about the values from various
pollutants in the areas in Skopje providing information about the
regions that have problems with air pollution.
Keywords— air quality, indicator, measurements, measurement
interpolation, ecoinformatics, open data, linked data
I. INTRODUCTION
Linked Open Data and Semantic Web principles are the
main contributors in realizing the idea of data reusability and
data scalability [1][2]. It gives the opportunity of linking the
information from various fields and enabling simple access to
them. With this approach, data becomes understandable not
only for humans, but also for computer machines [3].
On the other hand, the air pollution in Macedonia,
especially in its capital Skopje, is one of the biggest problems
that the citizens of Skopje have. It can be described as the
pollution of the atmosphere with gases, or dust of solid
materials, particulate matter as other substances whose
amounts are constantly increasing [4]. This information is
approved by the multiple measurement stations positioned on
several locations in Skopje and its environment. Some of them
offer public domains for data access.
In this paper, we demonstrate the process of transforming
the collected data from several measurement stations into
four- and five-star Linked Open Data. We aggregate all the
data collected from the services provided by stations of The
Ministry of environment and physical planning, The Institute
of Eco informatics at Faculty of Computer Science and
Engineering, and the CO2 measurements provided by the
project Skopje Green Route, into one centralized database.
They provide different air quality indicators. Afterwards, we
transform the collected data into four- and five-star data by
reusing the existing ontologies from the domain, necessary for
the transformation and annotation process and linking them to
the physical places where the measurements were taken [5].
The network of monitoring stations is very important in
urban environment because it provides information of actual
quantity of air quality indicators. The main problem is
impossibility of obtaining appropriate values for all points of
interest. There are several reasons and as most important we
can mention the price of the measurement stations.
Consequently, only the most important points could be
monitored. Air dispersions models are solving this problem by
providing estimations and predictions of the pollutants in the
air using mainly emissions and meteorological data. These
models include on mathematical algorithms based on
combinations of physical and chemical simulating the spread
of pollutants in the air. We describe the process of
interpolation of the air quality data obtained from the
measurement stations. This process of interpolation is
repeating on constant time intervals to obtain approximate
values of all air quality parameters [6]. The number of
measurement stations is upgradable and it is directly
proportional with the accuracy of the approximations.
In the final section, we propose and demonstrate several
use-case scenarios querying published data set and represent
the results on a heat map providing information about the
current pollution from various pollutants in the area of Skopje.
II. RELATED WORK
The problem with air pollution is emerging almost all
urban cities in the world. It is estimated that worldwide, 2
million people and more than half of them are in developing
countries, die every year from air pollution. By releasing of
the air indicators from measurement stations as open data
provide contribution of the public understanding and dialogue
around far-reaching and potentially data-rich aspects of life in
the city. Consequently, air quality measurement datasets are
already part of the LOD cloud. Home Weather ontology is
intended for weather phenomena and exterior conditions
providing property hasAirPollution which express an index of
air pollution depending on the current air quality measurement
values (Fig. 1).

Figure 1. HomeWeather dataset
AirQuality+ is a project
1
which gathers real-time air
quality measurements from different points in Sheffield,
England, providing open licenses for communities and
organizations to access and re-use. This includes near real-
time data on pollutant levels in Sheffield collected by a
network of monitoring stations, as well as data related to the
issue of air quality, such as industrial activity, traffic and
transport, public health, weather and land use.
The PESCaDO Ontology
2
is a modular application
ontology exploited for personalized environmental decision
support, that enables to formally describe:
• the user decision support request
• the environmental data relevant to process the request
• the decisions and conclusions to be produced
The PESCaDO Ontology was thoroughly developed
following state of the art best practices, and it is matched with
a comprehensive and detailed documentation.
Air dispersion models are based on mathematical
algorithms providing probabilistic values about the current air
quality at each point of the city. They are related to the city
infrastructure, the current weather conditions and the real-time
measurements from several measurement stations. Currently,
the Institute of Eco informatics has developed air dispersion
models about the capital of R. Macedonia, Skopje. They use
the measurements from The Ministry of environment and
physical planning providing interpolation values for each point
in the central region. These data is kept as 1 star data
providing interpolated monochromatic visualizations (Fig. 2)
on Skopje’s map, generated from ArcGIS server. In this paper,
we will convert the information from this visualization into 5
star data and we will provide useful information mining the
gathered dataset.
Figure 2. CO interpolated monochromatic visualization
1
http://betterwithdata.co/portfolio/air-quality-plus/
2
https://dkm.fbk.eu/technologies/pescado-ontology
III. DATA FROM SKOPJE’S AIR QUALITY MEASUREMENT
STATIONS
A. Centralizing data from multiple services
There are multiple measurement stations distributed in the
region of Skopje providing different air quality indicators.
Most of them, provide open access REST services which return
real-time measurements. In our paper, we will use the services
provided by:
• Ministry of environment and physical planning
The JSON service which provides measurement about CO,
NO2, PM10, PM2.5, SO2, O3 air quality indicators providing
data refresh each hour during the day. There are multiple
measurement stations over Macedonia, but for purposes of this
paper, we will use only the stations located in Skopje: Centar,
Karpos, Lisitche, Gazi Baba and Rektorat.
• Measurement CO2 stations provided by the project
Skopje Green Route
These measurement stations are placed on the most frequent
crossroads in Skopje: Justice Palace, Red Cross and Faculty of
Agriculture, providing measurement about CO2 air indicator
refreshing the information each 5 minutes.
• Measurement station maintained by the “Laboratory
of Eco informatics at Faculty of Computer Science and
Engineering” providing information about the same air
indicators like the measurements stations enabled by Ministry
of environment and physical planning
All services are RESTful and provide open URL location
which can be accessed with GET parameters. The log of all
services is kept on our database which centralizes all
information about all air quality indicators for all
measurement stations (Fig. 3). It runs scheduled processes
which poll the JSON services asking for new fresh data from
sensors. It appends timestamp to the measurement information
and saves in MySQL database whose EA diagram is
represented on Fig. 4.
Figure 3 Centralized data polling architecture.

Figure 4. EA diagram of the centralized database.
B. Interpolation of the measurements in the area of Skopje
The process of interpolation is based on the newest “up to
time” data as the average of the air parameter concentration
per hour for each parameter. The data are provided by the
network of measurement stations described in section A stored
in one centralized database. They are used for generation of
the grid raster layers by interpolating techniques. The model
depends on the weather conditions and the infrastructure of the
area taking as references the values from the nearest real
weather and pollution measurement stations. The interpolation
data is calculated on ArcGIS Simulation Server which
implements the pollution model [6], gets the measurements
from weather and pollution stations and provide interpolated
information about the area of Skopje. The output of the model
is a raster image, so we create algorithm for data
transformation from raster monochromatic image to numerical
format about the pollution state. The darker positions represent
greater values of pollution. After the transformation process,
we provide RESTful services which could be easily accessed
by setting the latitude and longitude of the required position.
This RESTful services is accessed by the interfaces of our
application. They append appropriate timestamp and current
weather conditions, and save in the centralized database.
C. PESCaDO Ontology
In order, to transform the measurement 3 star data, from
the centralized database, into RDF, we need ontology. Among
multiple ontologies that we reviewed in our research, the
PESCaDO Data ontology proved as the most useful for our
needs. It is developed by Data & Knowledge Management
research group [7] which is part of the Information and
Technology Center in Fondazione Bruno Kessler and it is
provided for mapping of measurement data from sensors of
PM10, PM2.5, CO and other air pollutants. It also provides
mapping of the weather conditions so we concluded that this
ontology is satisfying our needs.
We divided the properties of the ontology in two main
types: weather conditions properties (Table 1) and air
pollution indices (Table 2).
Table 1. Weather condition properties
Property
Type
Description
HumidityValue
Datatype Property
Air humidity value
TemperatureValue
Datatype Property
Ambient
Temperature value
WindSpeedValue
Datatype Property
Wind speed value
Table 2. Air pollution indices properties
Property
Type
Description
PM10IndexValue
Datatype
Provides information
about the value of the
PM10 particles in the
air
PM2.5IndexValue
Datatype
Provides information
about the value of the
PM2.5 particles in the
air
COIndexValue
Datatype
Provides information
about the value of the
CO concentration in
the air
NO2IndexValue
Datatype
Provides information
about the value of the
NO2 concentration in
the air
SO2IndexValue
Datatype
Provides information
about the value of the
SO2 concentration in
the air
O3IndexValue
Datatype
Provides information
about the value of the
O3 concentration in
the air
D. Geo Ontology
To provide mapping of the geographical location of the
measured or interpolated instance, we used the Geo ontology
which is one of the most used (Fig. 5). We have the correct
positions of the static measurement stations and we link the
measurement with the exact position of the station.
Afterwards, we divide the area of Skopje in zones providing
interpolated information about each zone separately. Each
zone has own latitude and longitude enabling linking to the
appropriate instance from the Geo ontology [8].
Figure 5. Geo ontology.
Table 3. Geo ontology
Property
Type
Description
Location
ObjectType
Description of the
geographic entity
Lat
Datatype
Latitude of the
mapped object
Lon
Datatype
Longitude of the
mapped object
E. Mapping the data from 3-star to 5-star data
After defining the ontologies, we need to transform the
data saved in database to RDF. In order to accomplish this, we

decided to use D2RQ server
3
which is compatible with
MySQL databases and provides accessing relational databases
as virtual, read-only RDF graphs without replicating into an
RDF store. Using D2RQ we can query a non-RDF database
using SPARQL, access to the content of the database as
Linked Data over the Web, create custom dumps of the
database in RDF formats for loading into the RDF store and
provides access to a non-RDF database using the Apache Jena
API.
The mapping process consisted of two steps. The first step
provides wrapping of the relational database, in our case
MySQL, with the interfaces provided by the D2RQ providing
access to the stored data. Afterwards, we should define a
mapping file using the D2RQ mapping language
4
to map the
relational database schemas to RDF vocabularies and OWL
ontologies. The mapping file defines a virtual RDF graph that
contains information about the database. This graph contains
RDF terms using d2rq:ClassMaps and d2rq:PropertyBridges.
The class map specifies how URIs (or blank nodes) are
generated for the instances of the class. It has a set of property
bridges, which specify how the properties of an instance are
created.
Our database, referencing to Figure 4, stores information
about Pollutant, Location and Measurement. As the image
represents, it is designed in 3rd normal form. As defined in
section D, we use PESCaDO and Geo ontologies so we need
to decompose the database in 2nd normal form providing the
table pollutant be part of the measurement. To solve this
problem, we change the mapping configuration using the
D2RQ mapping language, so we need not to make any
changes in the model of the relational database only by using
the property d2rq:condition. The property d2rq:condition
provides the SQL WHERE condition so an instance of this
class will only be generated for database rows that satisfy the
condition.
map:measurements_CO a d2rq:ClassMap;
d2rq:dataStorage map:database;
d2rq:uriPattern "measurement/@@T_MEASUREMENT.id@@";
d2rq:class pescadoData:COIndexValue;
d2rq:join "T_MEASUREMENT.pollutant_id =>
T_POLLUTANT.id";
d2rq:condition "T_MEASUREMENT.pollutant_id = 3";
d2rq:propertyDefinitionLabel "Measurements CO";
Transforming the data to 5 star data is provided by the
geo:Location property linking the measurement information to
specified Location where it is measured or interpolated. So, in
the global graph, we provide air pollution indicator for the
specified location.
F. USE CASE EXAMPLE
In this section we will demonstrate that transformation of
data into Linked Data, can provide useful use-case scenarios.
The result of use-cases gives opportunities for visual
3
http://d2rq.org/d2r-server
4
http://d2rq.org/d2rq-language
representation of the pollution on a heating map caused by all
pollutants separately.
By following query, we can obtain information about CO
measurements for the area of Skopje in determined time:
PREFIX rdf: <http://www.w3.org/1999/02/22-rdf-syntax-
ns#>
PREFIX pescado: https://ontohub.org/fois-ontology-
competition/PESCaDO_Ontology/pescadoData.owl#
PREFIX prov: <http://www.w3.org/ns/prov#> .
SELECT DISTINCT ?lat ?lng ?value WHERE {
?s rdf:type pescadoData:COIndexValue;
rdf:value ?value;
prov:atLocation ?location;
prov:generatedAtTime "2015-03-
03T19:15:46"^^xsd:dateTime.
?location geo:lat ?lat.
?location geo:lng ?lng.
}
This query starts executing over the local RDF graph
providing the measured and interpolated measurements in a
determined time from the area of Skopje. This query returns
similar data shown on the table 4:
Table 4. Partial result from the SPARQL query
Lat
Lng
Value
"42.05"^^xsd:fkiat
"21.32"^^xsd:fkiat
0.3
"41.96"^^xsd:fkiat
"21.31"^^xsd:fkiat
0.33
"41.94"^^xsd:fkiat
"21.29"^^xsd:fkiat
0.37
"42.03"^^xsd:fkiat
"21.3"^^xsd:fkiat
0.31
The result from the query could be used as input of a heat
map obtaining the visual representation of the concentration of
the CO in the air of the area of Skopje. The results from the
previous query are represented on Figure 5.
Figure 6. Visual heatmap representation of the CO measurement in the area
of Skopje.

Figure 7 Visual heatmap representation of the PM10 measurements in the
area of Skopje.
Analyzing the results on the heat map, we can conclude
that the municipality of Aerodrom has the highest values of
CO pollution.
The SPARQL endpoint for reviewing and analyzing the
results of measurements is available on the following url:
http://airpollution.b1.finki.ukim.mk/
IV. CONCLUSION
The concept of Linked Data represents a big advantage
in representation and retrieval of structured data from
distributed parts of the Web. A large number of communities,
companies and other interested stakeholders are taking part in
the initiative and are contributing to the expansion of the LOD
Cloud.
The type of the data that we contribute to open and to link
is providing interesting analyzes about the current state of the
air in the area of Skopje. We are allowing measurements about
CO, CO2, SO2, O3, NO2, PM10 and PM2.5 air pollutants. The
measurements are provided by 7 air pointer stations and 3 CO2
measurement stations. We provide interpolated values for the
areas that are not covered by the measurement stations.
Interpolated values are created by sophisticated models of
pollution spreading taking as parameters the infrastructure and
the model of spreading of the appropriate pollutant. We save
all of the data, measurements and interpolated values, in
centralized database that is wrapped by D2RQ server
providing mapping to RDF triples and linking to appropriate
locations in Skopje. We provide a URL for accessing the data
and reviewing the results using SPARQL query.
This type of data can help the citizens to find the best
places for their activities or for living in Skopje. Also it can be
used for retrieving the best eco routes for travelling in Skopje.
REFERENCES
[1] C. Bizer, T. Heath, K. Idehen, and T. Berners-Lee, “Linked data on the
web,” 17th International conference on World Wide Web, ACM, 2008,
pp. 1265-1266
[2] C. Bizer, T. Heath, and T. Berners-Lee, "Linked Data - the story so far,"
International Journal on Semantic Web and Information Systems 5, no.
3, 2009
[3] A. Naeve “The Human Semantic Web, Shifting From Knowledge Push
To Knowledge Pull”, International Journal on Semantic Web and
Information Systems (IJSWIS), 2005
[4] L. Barandovski, V. Urumov “Air pollution studies in Macedonia using
the moss biomonitoring technique, NAA, AAS and GIS technology”,
INIS, 2006
[5] M. Oprea “Mapping Ontologies in an Air Pollution Monitoring and
Control AgentBased System”, Lecture Notes in Computer Science
Volume 4265, 2006, pp 342-346.
[6] N.Koteli, K. Mitreski, D. Davcev “Monitoring, Modeling and
Visualization System of Traffic Air Pollution – A Case Study for the
City of Skopje”, International Conference on Mobile Ubiquitous
Computing, Systems, Services and Technologies, 2014
[7] V. Epitropou, L. Johanson, K.D. Karatzas, A. Bassouckos, A.
Karppinen, J. Kukkonen, M. Haakana, “Fusion of Environmental
Information for the Delivery of Orchestrated Services for the
Atmospheric Environment in the PESCaDO project”, International
Environmental Modelling and Software Society (iEMSs), 2012
[8] A. Patil, S. Oundhakar, A. Sheth, K. Verma, “Meteor-s web service
annotation framework”, Proceedings of the 13th international
conference on World Wide Web, pp 553 - 562