A Holistic, Semantics-aware Approach to Spatial Data Infrastructures

Cristiano Fugazza, Monica Pepe, Alessandro Oggioni, Fabio Pavesi and Paola Carrara

IREA-CNR, via Bassini 15, 20133 Milano, Italy

Keywords:

Spatial Data Infrastructures, Semantic Web, Marine Research, RDF.

Abstract:

We present a novel approach to the management of Spatial Data Infrastrutures that leverages semantics-aware

context information to model the distinct aspects involved in the management of geospatial data. RDF-based

schemata are employed for encoding information about the user community, the terminologies in use in a

speciﬁc research domain, gazetteer information representing the physical landscape underpinning data and,

last but not least, resource metadata. The data structures are then interconnected to enable seamless exploita-

tion for metadata creation and resource discovery, which we demonstrate through a worked-out example of

SPARQL query on RDF graph data. The methodology is being applied by the National Research Council of

Italy (CNR) to support creation of a distributed infrastructure for marine data in the context of the RITMARE

Flagship Project.

1 INTRODUCTION

Spatial Data Infrastructures (SDIs) represent one of

the most challenging areas in data management. In

fact, the variety of data and metadata formats, the dif-

ferent workﬂows in use, and the manifold expecta-

tions from the user community make it difﬁcult to ap-

ply widely acknowledged design patterns. Moreover,

the ever-changing landscape of data sources - think of

the emergence of sensor data (Na and Priest, 2007) -

and the evolution of data access methodologies - e.g.

the Linked Open Data (LOD) movement (Dodds and

Davis, 2012) - prevent data management practices to

easily coalesce into generally applicable standards.

Recourse to semantics in the management of

geospatial data is typically intended to address het-

erogeneity in resource description, to achieve multi-

lingualism in resource discovery functionalities, and

in general to compensate for the lack of the efﬁ-

cient search paradigms of generalized information re-

trieval. However, no existing infrastructure employs

semantics as the trait d’union between the distinct

components of the context information involved in

SDIs. Also, no state-of-the-art solution applies this

approach throughout the life cycle of geospatial in-

formation, from metadata production to resource re-

trieval. Instead, it is our opinion that only by adopt-

ing an all-round approach encompassing all the as-

pects of SDIs it is possible to fully exploit the po-

tential of the expressive descriptions enabled by on-

tologies, thesauri, and RDF-based data structures in

general

With this in mind, IREA-CNR

tackled the chal-

lenge of providing an infrastructure to the RITMARE

Flagship Project

, which aims at bringing all the dis-

tinct contributions to Italian marine research under the

same umbrella. In doing this, RDF-based data struc-

tures are exploited to model the distinct aspects of

the SDI. These range from the categorization of re-

searchers and institutions involved in the project to the

collection of the terminologies in use by the distinct

communities; from the organization of a gazetteer

system supporting geospatial information retrieval to

the encoding of metadata as Linked Data records. The

latter component will contribute to the LOD cloud

with a “bubble” constituted by the metadata collected

by RITMARE.

In this paper, we report on the activities that are

carried out for the establishment of the RITMARE

infrastructure with regard to semantics. Section 2

describes the project’s generalities and presents some

related work that is relevant to our research. Section 3

describes the RDF-based data structures that support

the front- and back-ofﬁce functionalities of the infras-

Resource Description Framework: http://www.w3.org/

RDF/

Institute for Electromagnetic Sensing of the Environ-

ment - National Research Council of Italy: http://irea.cnr.it/

RITMARE (Ricerca ITaliana per il MARE - Italian re-

search for the sea): http://www.ritmare.it/

Linked Open Data cloud: http://lod-cloud.net/

349

Fugazza C., Pepe M., Oggioni A., Pavesi F. and Carrara P..

A Holistic, Semantics-aware Approach to Spatial Data Infrastructures.

DOI: 10.5220/0004997603490356

In Proceedings of 3rd International Conference on Data Management Technologies and Applications (DATA-2014), pages 349-356

ISBN: 978-989-758-035-2

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

tructure. Section 4 provides a preliminary eval-

uation of the approach that is presented. Finally,

Section 5 draws conclusions and outlines future

work.

2 CONTEXT

The building of the data sharing infrastructure in RIT-

MARE (a Flagship Project by the Italian Ministero

dell’Istruzione, dell’Universit

a e della Ricerca) re-

quires integration of all state-of-the-art contributions

to Italian marine research into a coherent SDI. A

coarse-grained categorization of SDIs distinguishes

between centralized and decentralized infrastructures,

according to whether data and metadata is stored in

a single repository or distributed among the distinct

data providers. The RITMARE infrastructure belongs

to the second kind, featuring:

• a set of peripheral nodes that expose standards-

compliant metadata and services;

• a centralized catalog service that provides a single

point of access to the resources made available by

the project as a whole.

The project involves a heterogeneous set of data

providers (public research bodies and inter-university

consortia) as well as a variety of stakeholders (pub-

lic administrations, private companies, and citizens).

These entities envisage a varied corpus of heteroge-

neous data, metadata, workﬂows, and requirements.

Besides this, data providers feature different degrees

of maturity with regard to the provisioning of re-

sources according to the mandated standards. In fact,

the RITMARE SDI is bound to the requirements set

by the INSPIRE (INfrastructure for SPatial InfoR-

mation in Europe) Directive (European Commission,

2007), as well as by RNDT

, the Italian enforcement

of the former.

The development activity can be roughly divided

into three incremental phases. The ﬁrst one consists

of empowering the data providers with standards-

compliant services for the provisioning of geospatial

data. This has been achieved by providing a virtual

appliance, a “starter kit” (Fugazza et al., 2014), that

is capable of kickstarting an autonomous node in the

SDI for the collection, annotation, and deployment of

both geographic and sensor data. The data sources

that already expose standard services will be inte-

grated in the second phase. Finally, the third phase

will consist of the building of the centralized portal

Repertorio Nationale dei Dati Territoriali:

http://www.rndt.gov.it

integrating the distinct peripheral nodes and exposing

the metadata records as Linked Data.

In the ﬁrst two phases, the data structures de-

scribed in this work allow to weave into the workﬂow

for resource registration advanced functionalities for

the enrichment of metadata. In the third phase, the

centralized repository will exploit them for articulat-

ing resource discovery in a more effective way as re-

gards to existing geoportals (the front-end of SDIs).

2.1 Related Framework

INSPIRE-compliant, XML-based metadata and

OGC-based services (European Commission, 2008;

IOC Task Force for Network Services, 2011a; Initial

Operating Capability Task Force, 2012; IOC Task

Force for Network Services, 2011b) seem to ﬁnd it

hard to keep pace with more recent LOD initiatives,

even those sponsored by the EU itself, such as its

Open Data Portal

. However, an extension to the

DCAT

application proﬁle for data portals in Europe

(European Commission, 2013) is expected to ease

transition to RDF-based metadata representation

for INSPIRE metadata. The paper (Fugazza and

Luraschi, 2012) proposes an indexing process for

INSPIRE metadata based on SKOS thesauri. The

methodology is applied to harvested metadata records

and hence is likely to be error-prone, while our ap-

proach in this worst-case scenario (that is, when

harvesting metadata from existing infrastructures)

envisages supervised amelioration of metadata in

order to obtain RNDT-compliant metadata, providing

a more reliable indexing as by-product.

In extra-European contexts, such as the GEO Sys-

tem of Systems

, semantics is intended as the primary

means to overcome the generalized hetereogeneity of

formats (Khalsa et al., 2007; Nativi et al., 2011). In

particular, in (Santoro et al., 2012) the authors pro-

pose a query expansion mechanism that is applied to

the search string entered by the user, rather than to

the metadata records themselves. On the one hand,

this approach is valuable because it is applicable when

metadata records are not known in advance. On the

other hand, precision and recall in discovery is in-

evitably hampered by the scarce information associ-

ated with the user executing the search (e.g., the lan-

guage the query is expressed in is not known in ad-

vance).

Another important issue that we wanted to address

is the inclusion in our marine infrastructure of Sen-

EU Open Data Portal: https://open-data.europa.eu

Data Catalog Vocabulary (DCAT): http://www.w3.org/

TR/vocab-dcat/

GEOSS: https://www.earthobservations.org/geoss.shtml

DATA2014-3rdInternationalConferenceonDataManagementTechnologiesandApplications

350

sor Web Enablement

(SWE) functionalities, partic-

ularly the Sensor Observation Service (SOS) com-

ponent (Oggioni et al., 2014; Jiang et al., 2013).

Ontology-based representation of sensors and obser-

vations is featuring noteworthy results, as in (Bar-

naghia et al., 2011; Compton et al., 2012; Wang,

2012; Taylor et al., 2011; Taylor and Leidinger,

2011), that are inspiring our approach to the rep-

resentation of this category of data sources within

our RDF-based SDI. However, the challenge of har-

monizing sensor networks, particularly observations

(that is, the actual data produced by sensors), with tra-

ditional geographic data remains. Ideally, one should

transparently access sensor data alongside the other

categories of geospatial data and exploit the speciﬁci-

ties, characteristics, and history of sensors (e.g. mea-

sured parameters, gain, accurancy, offset, calibration

features) when needed. Instead, Sensor Web devel-

opment relies on speciﬁc standards and practices that

make it difﬁcult to achieve this degree of integration

(Havlik et al., 2011).

3 THE RITMARE LOD BUBBLE

Figure 1 depicts the individual components of the

RDF data structures that are being populated for de-

veloping the RITMARE infrastructure. The size of

each circle is proportional to the expected number of

entities in each data structure. Also, in order to exploit

these data structures in a seamless way, it is neces-

sary to provide the appropriate relations between en-

tities from distinct schemata. The following of this

Section will describe the speciﬁcities of the individ-

ual components with regard to i) the schemata that

are employed, ii) the data that are stored, and iii) the

relations that are created for connecting the distinct

domains. Section 3.5 provides some implementation

details on the two data bases that are conjunctively

employed. Finally, Section 3.6 describes an actual

discovery scenario, which is a worked-out example of

geospatial query interrogating all the data structures

to obtain a list of prospective results.

3.1 Project Description

Researchers and institutions involved in the project

are modeled as Friend Of A Friend (FOAF) data struc-

tures

. The rationale for this approach is the follow-

ing: Firstly, FOAF is a widely-acknowledged data

Sensor Web Enablement: http://www.opengeospatial.org/

ogc/markets-technologies/swe

FOAF Vocabulary Speciﬁcation 0.99: http://xmlns.com/

foaf/spec/

format for establishing a Web of Trust among the

participants to the project. Also, FOAF descriptions

can refer to terms in controlled vocabularies repre-

senting the research ﬁeld of individual persons and

whole institutions: As an example, Figure 1 shows

the FOAF entity corresponding to researcher “Mon-

ica Pepe” linked to the SKOS concept corresponding

to “Geophysics” by property topic interest from the

FOAF vocabulary. This semantic link is the primary

means we adopted to provide advanced metadata cre-

ation functionalities, enable proﬁled discovery of re-

sources, and tailor the selection of components to be

included in the infrastructure’s geoportal on the user’s

speciﬁcities.

The categorization of the entities involved in RIT-

MARE has been derived from documents provided

by the administration of the project as a spreadsheet.

The documents have been exported as XML and pro-

cessed with an XSLT stylesheet in order to obtain the

raw RDF/XML data structures. In this phase, human

supervision was required in order to prune duplicate

individuals, remove typos, etc. Finally, the workﬂow

for creating links between FOAF entities and terms of

the knowledge base is twofold: For testing purposes,

an ad-hoc backofﬁce application has been created; in-

stead, when the centralized geoportal will be avail-

able, the research ﬁeld associated with a given indi-

vidual will be one of the preferences that the user can

specify and modify.

3.2 Knowledge Base

The second component is constituted by a collection

of terminologies in the SKOS format. These include

the multilingual codelists that are employed in the cre-

ation of INSPIRE- and RNDT-compliant metadata,

a selection of thesauri whose terms can be used as

keywords, and the categorization of the RITMARE

research areas. An activity that may be required in

the future is the harmonization of distinct controlled

vocabularies, that is, the drawing of correspondences

between related terms in independent terminologies.

For instance, in Figure 1 the SKOS concepts corre-

sponding to “Elevation” and “Altitude” are related to

each other by semantic property closeMatch from the

SKOS ontology. This aspect is a fundamental en-

abling factor for query expansion functionalities that,

given the inter-disciplinary vocation of RITMARE,

may be an important component for the usability of

the infrastructure.

Whenever a suitable SKOS representation is avail-

able, thesauri have been downloaded from the author-

itative sources and inserted in the knowledge base

with little or no modiﬁcation. Instead, most of the

AHolistic,Semantics-awareApproachtoSpatialDataInfrastructures

351

Figure 1: Components of the RITMARE LOD bubble.

codelists that are used in INSPIRE metadata had to

be created from scratch by referring to INSPIRE and

ISO documentation. The same for the list of research

ﬁelds in RITMARE. As for the selection of thesauri

to be supported for keyword selection, consultation

with domain experts inside and outside of the project

recommended some of the thesauri provided by Sea-

DataNet

(particularly, those categorizing observa-

tion parameters, units of measure, sensor platforms,

and sensor instruments) and the GCMD Science Key-

words

provided by NASA. We deliberately avoided

including some thesauri that are typically used in the

geospatial domain, such as GEMET

, as they were

deemed as too general by this speciﬁc user commu-

nity. However, since referring to thesauri is a fun-

damental component in RITMARE metadata, we en-

courage the project partners to provide feedback on

new terminologies to be included in the knowledge

base.

SeaDataNet BODC webservices:

http://seadatanet.maris2.nl/v bodc vocab/welcome.aspx

Global Change Master Directory: http://gcmd.nasa.gov/

GEneral Multilingual Environmental Thesaurus:

http://www.eionet.europa.eu/gemet/

3.3 Gazetteer Information

Another adopted data structure is a ﬁne-grained rep-

resentation of the geographic features that may be re-

ferred to in user queries. For the time being, we only

included some of the toponyms provided by GeoN-

ames

in order to be able to express query patterns

like “100km West of Naples” and reduce recourse to

maps, bounding boxes, and other widgets that typi-

cally distinguish geographic data discovery from gen-

eralized search. Gazetteer data structures are already

referring to controlled vocabularies, also included in

the knowledge base, for categorizing toponyms. As

an example, Figure 1 shows two SKOS concepts (de-

scribed by literals “country, state, region, ...” and

“second-order administrative division”) that are re-

ferred to by the toponym corresponding to “Naples”

as values for, respectively, properties featureClass

and featureCode deﬁned by the GeoNames ontology.

In the future, it is possible that the GeoNames on-

tology could be complemented by (or directly repur-

posed for expressing) a custom RITMARE maritime

gazetteer.

The GeoNames RDF dump (around 3GB) is too

large for direct usage and too large for DOM and

GeoNames: http://www.geonames.org/

DATA2014-3rdInternationalConferenceonDataManagementTechnologiesandApplications

352

XSLT processing tools. We then developed a Python

script that extracted from the original ﬁle the Italian

toponyms. The thesaurus categorizing GeoNames’

toponyms also allowed to prune the toponyms that

were clearly of no interest to the project (e.g., those

corresponding to hotels, parks, etc.) so that the data

actually inserted in the triple store was just around

50MB.

3.4 Resource Metadata

Finally, the largest proportion of the RDF data em-

ployed by the RITMARE infrastracture is constituted

by resource metadata harvested from the peripheral

nodes. For their encoding, we rely on the Data Cat-

alog Vocabulary (DCAT) and its proﬁle tailored on

the requirements of the EU Open Data Portal. Un-

til a proper extension for expressing INSPIRE meta-

data is made available, we rely on the RDF repre-

sentation of ISO 19115 metadata provided by the

Australian CSIRO

. In the example in Figure 1, the

dataset “Dataset XYZ” has been related to the IN-

SPIRE Theme “Sea regions” and to the publisher

“Monica Pepe” by properties theme and publisher

from the DCAT vocabulary, respectively.

The enrichment of metadata records may follow

two distinct directions; in both cases, we essen-

tially annotate metadata items with the identiﬁers (the

URIs) of the RDF entities described above. When

these entities are already provided by sources in the

LOD cloud (e.g., GeoNames toponyms), they refer to

the corresponding Linked Data structures. The meta-

data produced via the tool that we developed in the

ﬁrst phase of the project (see Section 2) already con-

tain this information. Otherwise, when metadata is

harvested from an existing data source, indexing tech-

niques will be applied to complement the language-

dependent textual descriptions with the corresponding

language-neutral identiﬁers.

3.5 Implementation Details

The ﬁrst three components that are described above

are stored in an instance of the Virtuoso Universal

Server

(although only the triple store capabilities

are employed). This platform provides all the func-

tionalities that are required for the management of

RDF data and is accessed by the distinct components

of the architecture through the SPARQL

endpoint

ISO 19115 RDF representation:

http://def.seegrid.csiro.au/static/isotc211/iso19115/2003/

Virtuoso: http://virtuoso.openlinksw.com/

SPARQL Protocol And RDF Query Language:

http://www.w3.org/TR/rdf-sparql-protocol/

that is provided. The choice of the solution to be used

for the deployment of the fourth component, the meta-

data of the resources aggregated by the infrastructure,

is currently not deﬁnitive. For the time being, we are

evaluating CKAN

, the mainstream solution for the

implementation of open-data portals. Besides sup-

port for the provision of resource metadata as Linked

Data, this platform natively exploits the DCAT meta-

data format, which is the basis for the RDF-based re-

source description that we employ.

The motivations for using two distinct platforms

are manifold. Firstly, Virtuoso offers a wide set of

tools speciﬁcally aimed at database administration

and this feature has been dramatically important in

the beginning of the project, when multiple schemata,

thesauri, and prospective data representations were

evaluated. On the other hand, CKAN may constitute a

more scalable solution for the category of data that is

likely to grow in size, that is, resource metadata. Sec-

ondly, the distinct components may feature a different

status with regard to access control or authoritative-

ness of the source. For instance, user details involve

privacy issues that we wanted to keep under control.

Also, some components that constitute the knowledge

base (say, the SeaDataNet thesauri) already have their

authoritative sources and then providing them through

a second endpoint would be redundant. Finally, in

the future, some components of the context informa-

tion described in this Section, such as thesauri and

code lists, could be drawn directly from the respec-

tive SPARQL endpoints or Linked Data facilities. For

the time being, these data structures are materialized

in the local repository for performance and direct con-

trol.

3.6 Worked-out Example

The combined expressiveness of the four components

described above support the discovery capabilities ex-

empliﬁed in this Section, in which the metadata cat-

alog is interrogated by means of the SPARQL query

language. Queries are composed on the basis of pa-

rameters speciﬁed by the user through a discovery in-

terface (a web form) hinged on four elements:

WHO: This parameter is used for ranking results ac-

cording to the properties topic interest that are

speciﬁed for a given researcher (if authenticated).

The effect of this parameter is otherwise transpar-

ent to the end user.

WHAT: Describes the observation parameter of in-

terest to the user. When no match is found in the

CKAN The open source data portal software:

http://ckan.org/

AHolistic,Semantics-awareApproachtoSpatialDataInfrastructures

353

Listing 1: Example of SPARQL query for discovery of datasets.

1 PREFIX foaf : < http :// xmlns . com / foaf /0.1/ >

2 PREFIX vcrd : < http :// www . w3 . org / 2 0 0 6 / vcard / ns #>

3 PREFIX skos : < htt p :// w ww . w3 . org / 2 0 0 4/ 0 2 / s kos / co re # >

4 PREFIX dcat : < htt p :// w ww . w3 . org / ns / dc at # >

5 PREFIX dtc : < h t tp :// purl . org / dc / te rms / >

6 PREFIX ext : < h t tp :// def . s e e g r id . c s iro . au / isotc2 1 1 / i so 1 9 1 15 / 2 0 0 3 / extent # >

8 SELECT ? da t as e t ? title ? abstra c t

9 WH E R E {

10 ? dataset r df : type d cat : Datas e t ;

11 dct : tit l e ? title ;

12 dct : d e s c ri pt i on ? a b st r a c t ;

13 dca t : theme < http : // vocab . ner c . ac . uk / col l e c ti o n / P02 / c u r r e nt / PSAL /> ;

14 dca t : p u bl i s he r ? p ub l i sh e r .

15 ? publ i s h er vcrd : org < ht t p :// ri t m a r e . it / rd f d a ta / p r o je c t # ISMAR > .

16 {

17 ? dataset ext : s p a t ia lE xt en t ? bbox .

18 ? bb ox ext : no r th Bo un dLa t i tu de ? nl at ;

19 ext : s o ut hB ou ndL a t it ude ? sla t ;

20 ext : w e st Bo un dLo n g it ude ? wlong ;

21 ext : e a st Bo un dLo n g it ude ? elong .

22 FILTER ( (? nl at > 4 0. 9 6 4 2) && (? s lat < 4 0 .8 5 2 8 ) &&

23 (? w l o n g > 13.15 7 3 ) && (? e l o n g < 12.95 6 7 ) )

24 }

25 }

related controlled vocabulary, a free-text search is

triggered by the application.

WHERE: Speciﬁes the geographic context of the

search as a free-text parameter to be interpreted

by the application.

WHEN: Determines the temporal extent of interest

to the user.

Also, the user is given the option to reﬁne the search

parameters by specifying facet values for some meta-

data items. These are the ﬁelds whose possible values

are known in advance (such as the INSPIRE Theme,

the topic category, the language used in metadata,

etc.) or that can be statically computed and clustered.

In the example, the user has selected, in a text ﬁeld

that is autocompleted by leveraging the SeaDataNet

thesaurus for observation parameters, the term corre-

sponding to “Salinity of the water column”. She also

speciﬁed the free-text value “100km west of Naples”

as the geographic location of interest. It is up to the

application to guess (in a Google Maps-style) the dis-

tance, direction, and origin expressed in the text ﬁeld.

Moreover, she also selected, among the data providers

in project RITMARE, the facet value corresponding

to the “ISMAR” institute.

Executing SPARQL queries amounts to searching

the graph data whose components have been intro-

duced in Section 3 for triple patterns. These are se-

quences of subject-property-object patterns (subjects

and objects like the ellipses in Figure 1, the proper-

ties like the directed arcs) inducing a graph that can

then be matched against the data graph. User queries

are translated into SPARQL queries: In particular, the

parameters described above translate into Listing 1.

The explanation of the listing is the following.

Lines 1-6 deﬁne namespace preﬁxes that allow to

shorten the triple patterns in the listing. Namely, the

preﬁxes identify the namespaces of the following vo-

cabularies: FOAF, vCard (used in FOAF), SKOS,

DCAT, and Dublin Core Terms (used in DCAT). Also,

preﬁx ext identiﬁes the data structures that compen-

sate for the lack of proper DCAT data structures for

expressing some INSPIRE metadata items. The SE-

LECT clause on line 8 deﬁnes the three variables that

shall be returned by the query, that is, the URI, ti-

tle, and abstract of the resources matching the query.

The WHERE clause deﬁnes the triple patterns that are

searched in the RDF data structures of RITMARE and

which data shall be bound to variables. Lines 10-12

retrieve the data requested in the projection clause on

line 8, while lines 13-15 is where the actual match-

ing starts. Speciﬁcally, the ﬁrst line matches metadata

records that are related to the observation parameter

that has been requested by the user. The remaining

two require that the point of contact that is provided

for data belongs to the institute that has been selected

by the user for the corresponding facet.

Finally, lines 17-23 express the tricky part of the

query, that is, the matching of the bounding box spec-

iﬁed by the user with the free-text parameter “100km

DATA2014-3rdInternationalConferenceonDataManagementTechnologiesandApplications

354

west of Naples”. Here, the geographic coordinates

corresponding to Naples (40.88333N, 14.41667E ac-

cording to the toponym speciﬁed by GeoNames) are

applied the offset corresponding to 100km westwise

(obtaining 40.90802N, 13.05750E). Then, a bound-

ing box has been computed from this position (the

four values in the FILTER clause on lines 22 and 23

of Listing 1). Here, we assume that the user is ﬁne

with the default dimensions of the bounding box pro-

vided by the discovery application, 16km wide and

11km high. The (possibly multiple) geographic ex-

tents of each candidate dataset are matched against

the search bounding box. Here, we assume the more

conservative approach, requiring the latter to be fully

contained in the former. However, a range of different

interpretations are possible. Particularly, in the future

we will aim at providing the user with the capability

of modulating the degreee of overlapping of bounding

boxes according to the number of results that would

be returned by the query.

4 DISCUSSION

In the current phase of the project, when the cen-

tralized geoportal is still work in progress, an overall

evaluation of the proposed solution is not easy. How-

ever, we are already collecting feedback on the ca-

pacity building of data providers envisaged in the ﬁrst

phase of the project (see Section 2), that is, we can

report on the efﬁcacy of our “starter kit”. In particu-

lar, we can evaluate the features that we are providing

for the assisted editing of metadata, which are heav-

ily dependent on the data structures described in this

paper.

With respect to these, recourse to the RDF data

structures presented in Section 3 allows to dramati-

cally reduce the number of the data items to be ac-

tually provided by the user. As an example, the

RNDT speciﬁcation is featuring 34 mandatory meta-

data items, too high a number for an accurate editing

of metadata records. Moreover, some of these, such as

the 4 distinct contact points to be provided, are com-

posite items that require more than one value to be

entered by the user. Instead, by combining i) the auto-

matic assignment of metadata items, ii) the provision

of default values extracted from the dataset itself, and

iii) the exploitation of the RDF information described

in this paper, the number of required values can be

narrowed down to 10.

Particularly, by referring to the project structure,

the editing interface is capable of providing data

items that are otherwise tedious to provide or error-

prone. Also, autocompleting keywords from the se-

lected thesauri conﬁgures a win-win situation where

the user is relieved of the burden of writing key-

words in their entirety and the application is capable

of associating unique identiﬁers with the keywords

that are provided. These advanced editing capabili-

ties can be applied to generic metadata schemata en-

coded as XML. In fact, we currently support the edit-

ing of sensor metadata in the SensorML format with

the same autocompletion facilities of RNDT meta-

data. Also, support for sensor metadata in the alterna-

tive *FL (Starﬁsh Fungus Language) format (Simonis

and Malewski, 2011) is under testing.

5 CONCLUSIONS

This work presents our approach to semantic charac-

terization of the data structures that are involved in

the management of SDIs. We believe that, by creating

the appropriate relations between data items encoded

according to the heterogeneous RDF schemata pre-

sented in Section 3, it is possible to provide end users

with advanced functionalities. These span from back-

to front-ofﬁce capabilities, from metadata creation by

the peripheral nodes of the RITMARE infrastructure

to resource discovery by the centralized catalog, and

permeate the whole life cycle of geospatial resources

as semantic annotations to INSPIRE-compliant meta-

data.

The proposed methodology is supporting the de-

velopment of the user-oriented infrastructure of the

RITMARE Flagship Project and constitutes the ﬁrst

example of decentralized management of Italian ma-

rine research data. The novelty of the approach also

resides in the all-round application of semantics in

the modeling of the data structures that describe re-

sources, end users, research ﬁelds, and physical lo-

cations. The example provided in Section 3.6 only

scratches the surface of the possible applications of

this multi-tiered knowledge base in the enactment of

Spatial Data Infrastructures.

ACKNOWLEDGEMENTS

The activities described in this paper have been

funded by the Italian Flagship Project RITMARE.

REFERENCES

Barnaghia, P., Ganza, F., and Abangara, H. (2011).

Sense2Web: A Linked Data Platform for Semantic

AHolistic,Semantics-awareApproachtoSpatialDataInfrastructures

355

Sensor Networks. semantic-web-journal.net, 0(0):1–

11.

Compton, M., Barnaghi, P., Bermudez, L., Garca-Castro,

R., Corcho, O., Cox, S., Graybeal, J., Hauswirth, M.,

Henson, C., Herzog, A., Huang, V., Janowicz, K.,

Kelsey, W. D., Phuoc, D. L., Lefort, L., Leggieri,

M., Neuhaus, H., Nikolov, A., Page, K., Passant, A.,

Sheth, A., and Taylor, K. (2012). The SSN ontology

of the W3C semantic sensor network incubator group.

Web Semantics: Science, Services and Agents on the

World Wide Web, 17(0):25 – 32.

Dodds, L. and Davis, I. (2012). Linked Data Patterns,

http://patterns.dataincubator.org/book/.

European Commission (2007). “Establishing an Infrastruc-

ture for Spatial Information in the European Commu-

nity (INSPIRE)” Directive 2007/2/EC.

European Commission (2008). Commission regulation

(EC) no 1205/2008 of 3 December 2008 implement-

ing Directive 2007/2/ec of the European Parliament

and of the Council as regards metadata.

European Commission (2013). DCAT Appli-

cation Proﬁle for data portals in europe,

https://joinup.ec.europa.eu/asset/dcat application pro

ﬁle/home.

Fugazza, C. and Luraschi, G. (2012). Semantics-Aware

Indexing of Geospatial Resources Based on Multilin-

gual Thesauri : Methodology and Preliminary Results.

International Journal of Spatial Data Infrastructures

Research, 7:16–37.

Fugazza, C., Menegon, S., Pepe, M., Oggioni, A., and Car-

rara, P. (2014). The RITMARE Starter Kit: Bottom-

up capacity building for geospatial data providers. In

9th International Conference on Software Paradigm

Trends (ICSOFT-PT).

Havlik, D., Schade, S., Sabeur, Z. a., Mazzetti, P., Watson,

K., Berre, A. J., and Mon, J. L. (2011). From Sen-

sor to Observation Web with environmental enablers

in the Future Internet. Sensors (Basel, Switzerland),

11(4):3874–3907.

Initial Operating Capability Task Force (2012). Technical

guidance for the implementation of inspire download

services.

IOC Task Force for Network Services (2011a). Technical

guidance for the implementation of inspire discovery

services.

IOC Task Force for Network Services (2011b). Technical

guidance for the implementation of inspire view ser-

vices.

Jiang, Y., Guo, Z., Hu, K., Shen, F., and Hong, F. (2013).

Using Sensor Web to Sharing Data of Ocean Observ-

ing Systems. In Wang, R. and Xiao, F., editors, Ad-

vances in Wireless Sensor Networks, pages 137–156.

Springer-Verlag Berlin Heidelberg, Berlin Heideberg.

Khalsa, S. J. S., Nativi, S., Ahern, T., Shibasaki, R., and

Thomas, D. (2007). The GEOSS interoperability pro-

cess pilot project. In IGARSS, pages 293–296.

Na, A. and Priest, M. (2007). OGC Sensor Observation

Service (OGC 06-009r6). OpenGIS Implementation

Standard.

Nativi, S., Craglia, M., Vaccari, L., Santoro, M., and

Fugazza, C. (2011). Searching the New Grail: Inter-

Disciplinary Interoperability. In Claramunt, C., Lev-

ashkin, S., and Bertolotto, M., editors, GeoSpatial Se-

mantics, Proceedings of the 4th International Confer-

ence, Geos 2011, number 1, pages 1–15. Springer-

Verlag.

Oggioni, A., Bastianini, M., Carrara, P., Minuzzo, T., and

Pavesi, F. (2014). Sensing Real-time Observatories

in Marine Sites A Proof-of-Concept. In SensorNets

2014, 3rd International Conference on Sensor Net-

works, pages 111–118, Lisbon Portugal 7-9 January

2014.

Santoro, M., Mazzetti, P., Nativi, S., Fugazza, C., Granell,

C., and D

ıaz, L. (2012). Methodologies for aug-

mented discovery of geospatial resources. In D

ıaz,

L., Granell, C., and Huerta, J., editors, Discovery of

Geospatial Resources: Methodologies, Technologies,

and Emergent Applications, chapter 9.

Simonis, I. and Malewski, C. (2011). *FL Starﬁsh Fungus

Language for sensor description. OGC discussion pa-

per 11-058r1.

Taylor, K., Compton, M., and Laurent, L. (2011).

Semantically-enabling the web of things: The W3C

semantic sensor network ontology. In Proc. of 5th

eResearch Australasia Conference.

Taylor, K. and Leidinger, L. (2011). Ontology-driven com-

plex event processing in heterogeneous sensor net-

works. In Antoniou, G., Grobelnik, M., Simperl, E.,

Parsia, B., Plexousakis, D., Leenheer, P., and Pan,

J., editors, The Semanic Web: Research and Appli-

cations, volume 6644 of Lecture Notes in Computer

Science, pages 285–299. Springer Berlin Heidelberg.

Wang, P. (2012). Semantically enabling next generation en-

vironmental informatics portals. PhD thesis, Rensse-

laer Polytechnic Institute.

DATA2014-3rdInternationalConferenceonDataManagementTechnologiesandApplications

356