Computers & Geosciences 36 (2010) 10691080

Contents lists available at ScienceDirect

Computers & Geosciences

journal homepage: www.elsevier.com/locate/cageo

Interactive visualization of marine pollution monitoring and forecasting data

via a Web-based GIS
M. Kulawiak a,n, A. Prospathopoulos b, L. Perivoliotis b, M. uba a, S. Kioroglou b, A. Stepnowski a
a
b

Gdansk University of Technology, Department of Geoinformatics, Gabriela Narutowicza 11/12, 80-233 Gdansk, Poland
Hellenic Centre for Marine Research, Institute of Oceanography, 46.7 km Athens-Sounio Ave., P.O. Box 712, 19013 Anavyssos, Greece

a r t i c l e in fo

abstract

Article history:
Received 27 October 2009
Received in revised form
11 February 2010
Accepted 15 February 2010

This work implements a Web-based Geographic Information System (Web GIS) for an existing oil spill
monitoring and forecasting service, developed in the framework of the MARCOAST project. This is
achieved by remotely presenting the results of the oil spill forecasting module via a dedicated Web GIS,
which allows authenticated end users to view the simulation results in a geographical context. A
number of Web GIS technologies for presentation of dissimilar types of semi-dynamic geographic data
are applied and their respective capabilities for publishing and remote presentation of interactive
geospatial information, such as oil spill spread animation overlaid on background data (terrain
elevation data, satellite imagery, etc.), are described. More specically, technologies like ESRIs ArcIMSs
(Arc Internet Map Server) and Open Source GeoServer with OpenLayers client library are implemented.
The capabilities of the system for visualization and mapping are illustrated by specic applications
concerning the spreading scenarios of oil spills in two selected areas of the Aegean Sea, Greece. The
presented Web GIS offers added value in the form of providing the end user with comprehensive and
synthetic, both spatial and temporal, environmental information through a remotely customizable
user-friendly graphical interface. In this context, its integration to a marine pollution monitoring and
forecasting system could result in an enhanced pollution awareness and emergency management tool.
& 2010 Elsevier Ltd. All rights reserved.

Keywords:
Oil spill prediction
Operational system
Web GIS
Open Source
GeoServer
Environmental management

1. Introduction
The increasing demand for overseas oil transportation puts the
marine environment under considerable pressure. Ecosystem
contamination caused by crude oil leaks, either accidental or
intentional, can be particularly dangerous when the geographical
context renders the marine environment very sensitive to
increase in pollution, e.g. in semi-enclosed basins like the Aegean
and Baltic Seas. A representative work (Pavlakis et al., 2001),
based on spaceborne synthetic aperture radar (SAR) remote
sensing, reveals the dramatic dimension of shipping pollution in
the Mediterranean Sea caused by routine unauthorized operational discharges. Therefore, strategic planning and preparation of
appropriate response scenarios to possible accidents is required
by most maritime district authorities. In this connection, the
development of operational systems and services for monitoring
and forecasting of pollution in the marine environment, which
provide near real-time information in a user-friendly way to the
appropriate authorities, is almost a compulsory need. This, in
turn, calls for cooperation of reliable oil spill simulation models

Corresponding author. Tel.: +48 58 3471728; fax: + 48 58 3472090.

E-mail address: Marcin.Kulawiak@eti.pg.gda.pl (M. Kulawiak).

0098-3004/$ - see front matter & 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.cageo.2010.02.008

and modules for illustrating, visualization and dissemination of

the results.
The considerable development of Web-based Geographic
Information Systems (Web GIS), which allow for creation,
publishing and analysis of geospatial data through the internet
without a traditional desktop GIS application, seems to provide
the means towards the aforementioned targets. The potential of
Web-based mapping and virtual reality technologies for environmental modelling has been considered about a decade ago (Doyle
et al., 1998). Within this decade, Web GIS related research and its
applications have been making rapid progress. Computers with
fast processing units and large amounts of memory allow for
practical applications of time consuming and resource intensive
algorithms, both for spatial analysis as well as visualization of
large-volume data, such as high resolution raster images, digital
terrain model (DTM) and vector data. The simultaneous progress
of Web mapping technologies has allowed for dissemination of
centrally stored geographic data and georeferenced remotely
sensed images via the World Wide Web (Su et al., 2000). However,
most Web GISs are used for presentation of static data, while the
area of dynamic and semi-dynamic data visualization has been
mostly neglected (Kraak, 2004). Novel applications of known
devices and models, such as SAR and oil spill simulations, require
interactive data presentation, which up to now has been mostly

1070

M. Kulawiak et al. / Computers & Geosciences 36 (2010) 10691080

implemented using standalone applications or thick web clients.

Some recent progresses on Web GIS environmental applications
are presented in the following characteristic works. The development of a Web-based hydrological modelling system that permits
integrated handling of real-time rainfall data from a wireless
monitoring network is presented by Al-Sabhan et al. (2003). In the
work of Sugumaran et al. (2004), the development of a Web-based
Decision Support System (DSS) for identifying and prioritization
of local watershed environmental sensitivity was described, using
ArcView Internet Map Server technology and a simple model. The
integration of ArcIMSs into a system for dissemination of data on
the location of boreholes, accompanied by strata information and
results of in situ and laboratory tests, has been presented by
Chang and Park (2004). The potential and opportunities afforded
by the integration of Web GIS and on-line remote sensing
facilities for environmental monitoring and management are
demonstrated by Tsou (2004), where open standards allow for
interconnecting different web mapping technologies that used to
operate on incompatible protocols. In (Kontoes et al., 2005), a
Web GIS tool has been implemented in order to provide the users
with specic knowledge on SAR image availability over a target
area and assess the monitoring capability with respect to the
requirements of oil spill detection over the SE part of the
Mediterranean Sea. More recently, Rao et al. (2007) implemented
a Web GIS DSS for use in resource management and assessment of
environmental quality, using web server and Java Servlet
technology over an ArcIMSs platform, while Kruger et al. (2007)
demonstrated a Web GIS built with MapServer, PHP and Javascript technologies, dedicated for verication of hydrologic
forecasts. In the same year, Tuama and Hamre (2007) presented
a Web-based DSS for marine pollution monitoring, integration
and distribution of multi-source measured data and results were
obtained from numerical models by implementing Web GIS
interface as a thin client and using non-proprietary technologies
and open standards. This method of GIS development has proven
to be successful and is being continuously utilized for implementation of similar projects, as shown by Hamre et al. (2009).
The latter work presents systems, products and services developed in the FP6 InterRisk project along with an outline of
technical development of the relative Web GIS client and portal
structure, focusing on the capability of the systems to integrate
and provide data from various sources.
The presented work concentrates on the use of Web GIS
technologies for linking a complex operational oil spill (or, in
general, environmental) forecasting model with a thin Web GIS
client in order to provide the end user with animated results of
the oil spill dispersion forecasts together with other valuable,
both temporal and spatial, information (e.g. oil spill characteristics for each simulation step and locations of particular
signicance such as specially protected or densely populated
sites) for supporting the corresponding risk assessment and
management. More specically, the present study, carried out in
the framework of a bilateral Polish-Greek R&T collaboration
between Gdansk University of Technology (GUT) and Hellenic
Centre for Marine Research (HCMR), aims to integrate a Web GIS
to an existing oil spill monitoring and forecasting service,
developed in the framework of the MARCOAST project. Its main
concept has been described by Prospathopoulos et al. (2008). The
forecasting module of the system, developed by HCMR in the
framework of the POSEIDON project, consists of a complex oil spill
model, designed to use the results of three operational models
(meteorological, waves and hydrodynamic) that on a daily basis
provide 72 h forecasts for the Aegean and the Eastern Mediterranean Sea. The use of accurate information on atmospheric and
oceanic conditions results in a more realistic description of the
fate of oil in the sea, compared to applications, which use

simplied approaches for the air/sea conditions. The results and

relative information obtained from the oil spill forecasting system
are remotely presented by means of a dedicated Web GIS,
developed by GUT. The Web GIS allows authenticated end users
to view the simulation results in a geographical context, in the
form of various thematic layers overlaid on background marine
electronic charts and land data. The possibilities of presenting
semi-dynamic geographic information and publishing interactive
geospatial data, such as animation of oil spill spread overlaid on
background data, through the use of Web GIS technologies, like
ESRIs ArcIMSs and Open Source GeoServer with OpenLayers
client library, are considered. The capability of the dedicated Web
GIS module for visualization and mapping is illustrated by specic
applications concerning two different events of oil spill detection
and spreading by the MARCOAST oil spill monitoring and
forecasting service in two areas of the Aegean Sea, Greece.

2. The marine pollution monitoring and forecasting system

Although the Web GIS developments presented in this work
have taken into account the possibility of visualizing and mapping
data from different sources as outputs of an integrated marine
pollution monitoring and forecasting system, oil spill dispersion
data provided by an existing operating system was chosen for
testing the performance of the aforementioned developments.
Thus, the Web GIS input was obtained from the standard output
of an integrated oil spill monitoring and forecasting service, which
is delivered as one of the core downstream services of MARCOAST
project in the area of the Aegean Sea. The structure and the
components of this service, briey described in the Sections 2.1
and 2.2, are depicted in Fig. 1.
Based on the information provided by the aforementioned
service, two oil spills had been detected on 13 April 2007,
8:27 UTC, at (26.0401E, 39.6641N) north of Lesvos Isl., and at
(23.8191E, 37.5911N), southwest of Cape Sounio. The corresponding spreading events were selected as pilot events for testing the
performance of the Web GIS developments in this work. The
initial locations of the detected oil spills and the wider areas
boxing them are shown in Fig. 2.
2.1. The MARCOAST oil spill monitoring and forecasting service
The European Space Agency (ESA) funded project Marine and
Coastal Environmental Information Services (MARCOAST, 2005
2008) belongs to the European action known as Global Monitoring
for Environment and Security (GMES). This effort is co-funded by
the ESA and the European Union and aims to deliver a DSS for use
by the public and the policy makers, with the capability of
acquiring, processing, interpreting and distributing information
related to environment, risk management and natural resources.
In this framework, MARCOAST aims to deliver a single portfolio of
the satellite-based services in the eld of marine and coastal
applications at European scale. The provided services integrate
detection and monitoring technologies involved in water quality,
oil spill and meteorological information into a durable network.
The MARCOAST portfolio is composed of six main service lines:
Oil spill surveillance and customized information (S1), Oil spill
drift forecast (S2), water quality monitoring and alert (S3),
harmful algae bloom monitoring, evolution and forecasting (S4),
water quality assessment service (S5), and met-ocean data (S6). In
the area of the Aegean Sea, the implemented service is a
combination of the oil spill detection processes, applied on
satellite-based SAR images, together with state-of-the-art numerical models that are able to forecast the oil spill evolution in the
marine environment. The deployment of the service is a joint

M. Kulawiak et al. / Computers & Geosciences 36 (2010) 10691080

1071

FORECASTING (F)
(POSEIDON Operation Tool for oil spill prediction)
Forcing module

MONITORING (M)

Input

Processing

Output (M)

3D oil spill
weathering &
drift module

Meteorological model

Wind-Wave model
DISSEMINATION

Output (F)

3D Hydrodynamic
model

Fig. 1. Structure and components of MARCOAST oil spill monitoring and forecasting service.

Fig. 2. Areas of pilot implementation of Web GIS in Aegean Sea (indicated with boxes) and initial locations of detected oil spills (indicated with pins).

effort of Telespazio (Rome, Italy), which receives and analyses the

SAR images, and the Hellenic Centre for Marine Research (HCMR),
which implements the numerical models to predict the fate of the
oil in the sea. The service is designed to deliver, in less than 1 h
after the satellite data acquisition, a web-based report concerning
the oil spill detection and its forecasted fate. The Hellenic Ministry
of Mercantile Marine, which is the responsible authority in Greece
for the marine environment surveillance, is alerted by fax/e-mail/
SMS to visit the services dedicated website and receive all
relevant information about the recorded event.
The oil spill service for the Aegean Sea is made up by the
following three components:
(1) Monitoring module: Routine acquisition and analysis of the
SAR images taken by ENVISAT satellite over the Aegean Sea.

This process is realized by Telespazio (Italy). Briey, the

monitoring system consists of:
 Input data (raw data from Envisat ASAR wide swath imagery,
downlinked directly to Telespazio ground station located in
Matera),
 Processing procedure (images are geocoded and assimilated
into oil slick identication model within 1 h), and
 Output data (slick location, extension, barycentre, surface,
length and width in ASCII and shape le formats as well as a
simple map with the target location in a georeferenced grid,
sent to HCMR).
(2) Forecasting part: The POSEIDON operational tool for the
prediction of oating pollutant transport, described in the
next subsection, is used. The oil spill drift forecasts are
provided in terms of sequential graphs showing the oil spill

1072

M. Kulawiak et al. / Computers & Geosciences 36 (2010) 10691080

evolution every hour for the next 72 h (maximum) after the

time of event. In each graph, the user can nd information
about the volume of the oil that has been evaporated,
emulsied, reached the sea oor or impacted on the
shoreline.
(3) Dissemination part: A dedicated web page (http://marcoast.
hcmr.gr) is created to host the detailed information about the
oil spill detection and forecasting. The access to this page is
restricted to registered users only. The end user is alerted by
e-mail/fax/telephone of new information posted to the
website.
2.2. The POSEIDON operational tool for oil spill prediction
The operational tool for prediction of oating pollutant
transport of the POSEIDON system (Pollani et al., 2001) is used
for oil spill prediction. This tool consists of two discrete modules:
a three-dimensional oil-spill weathering and drift module that
simulates the dispersion of oil droplets and their chemical
transformations, and a forcing module that provides data for
wind, waves, currents and diffusivity to the oildrift module.
The oil-spill weathering and drift module is based on PARCEL
model, see (Petihakis et al., 2001), which is able to simulate not
only the drift of the oil but also the chemical transformations
under the specic environmental conditions. The oil slick is
represented as parcels that have time dependent chemical and
physical characteristics. The basic processes simulated by the oilspill model are evaporation, emulsication, beaching and sedimentation. The evaporation process illustrates the transfer of oil
from the sea surface to the atmosphere, which can result in 20
40% loss of oil in the rst few hours. The emulsication process
describes the mixing of water in the heavier fractions of the
hydrocarbons. The sedimentation process describes the trapping
of oil particles that reach the sea bottom, while the beaching
process addresses the trapping of oil along the coast depending on
the type of coastline (rocky cliff, tidal ats, etc.). Further
description on how these processes are represented in the
numerical model can be found in (Petihakis et al., 2001), (Pollani
et al., 2001), (Pollani, 2005) and (Nittis et al., 2006).
The oil-spill weathering and drift module requires the
following basic input: initial location and shape/size of the oil
spill (latitude, longitude), date and time of the event, and an
estimation of the total volume of oil disposed in the sea.
Additionally, if available, information about the type of oil can
also be used. The main output of the drift model can be
summarized as follows: longitude, latitude and depth of each
particle in the sea, evaporated volume of the initial oil,
emulsicated volume, volume of the oil that reaches/remains on
the beach and the volume of oil that reaches the sea oor.
The basic background information (forcing module) for the oil
spill forecasts (wind, waves and 3-D currents) is currently
obtained from the forecasting products of the POSEIDON operational oceanographic system, installed and operated in the Greek
Seas since 1999 (Soukissian et al., 2000). In order to provide
realistic forcing data to the oil spill module, the forcing module
uses interpolated elds from three different operational models of
the system, which provide every day 72 h forecasts for the Aegean
and the Eastern Mediterranean Sea:
(1) The meteorological forecasting model, which is based on the
ETA Limited Area Model (Papadopoulos et al., 2002) and
downscales the global NCEP or ECMWF forecasts over the
Mediterranean Sea (0.241 grid) with a high resolution nesting
over the Aegean Sea (0.11 grid). This model provides airsea
boundary conditions to the ocean models.

(2) The windwave model based on the WAM model (The

WAMDI Group, 1988) provides wave height and direction
information for estimating the Stokes drift.
(3) The 3-D hydrodynamic model based on the Princeton Ocean
Model (Blumberg and Mellor, 1987) provides the three
components of the current eld as well as the horizontal
and vertical diffusivities.
3. Specications of the Web-based GIS
The Web GIS for visualization and mapping of the results from
the previously described monitoring and forecasting system was
designed and built to meet a set of requirements, which can be
divided into three categories:
(1) Input data. As mentioned in the previous section, the
MARCOAST output data are ASCII les containing simulation
results for an oil spill spread during a given amount of time.
Every oil spill is represented as a set of thousand geographically oriented droplets and is accompanied by additional data,
such as the simulated date and time, the total volume of the
oil spill as well as the evaporated and emulsied volumes
both in the sea and on land. All these informations were to be
extracted and processed by the Web GIS.
(2) Visualization. Authenticated end users were to view the
extracted information in a geographical context, in the form
of thematic layers overlaid on background sea and land data.
In addition to standard GIS operations, like view zooming and
panning, the system had to present to the user an animation
of the oil spill spread as well as tools for querying each hour
layer for parameters like the volume of selected oil spill, its
geographic coordinates, depth of its elements, etc. The
remaining requirements for the module encompassed multilayer visualization and mapping, integration of marine data
from different sources and Internet access through a standard
Web browser.
(3) User-specic. Taking into account that the system was
designated for use by municipal ofcers and coastal guards,
it had to be designed with great ease of use in mind. One of
the associated restrictions was to provide a DHTML client
independent from third-party technologies like Java, Flash,.
NET etc., which would force the user to download and install
separate plugins.
4. Web-based GIS architecture, capabilities and performance
Because the oil spill data needed to be processed in one place
and later be presented to users in several different places, it was a
logical decision to split these functionalities of the Web GIS
between a server and a client. Furthermore, because the end users
would not be engineers, the client needed to require a possibly
small amount of technical knowledge to handle. This clearly
indicated a clientserver architecture with a thin and possibly
lightweight client. The resulting high-level concept of the Web
GIS architecture is shown in Fig. 3.
According to this concept, the data presentation module would
consist of four main elements: the Data conversion module, the
Data registration module, the Map Server and the DHTML Client.
The output data from the monitoring and forecasting modules of
the MARCOAST system would be received by the data conversion
module, where it would be translated into a format compatible
with the map server. The data registration module would be
responsible for inserting the new data into the map server
database, where it would undergo reprojection and become
integrated with background data layers. Finally, the data would

M. Kulawiak et al. / Computers & Geosciences 36 (2010) 10691080

1073

Fig. 3. Concept architecture of Web GIS module.

become available for interactive visualization by remote users via

the DHTML Client.
The next step in building the data presentation module
involved the choice of appropriate solutions for implementing
the Web GIS sub-modules. The determining factor in the selection
of Web technologies used for building most sub-modules was the
choice of software for the map server module. The natural rst
choice was industry-standard ESRIs software, as both of the
development teams had good experience working with the
ArcIMSs.

4.1. Implementation of the ArcIMSs option to oil spill data

When the project began, the ESRIs ArcIMSs, now a discontinued product, was one of the most advanced tools for publishing
maps in the Web. Its capabilities encompass rendering of
multilayer raster and vector map images, which allow for remote
visualization on thin Web clients, as well as image and feature
streaming, which allows for integration of server-side data with
client-side layers in ArcGISs Desktop applications. ArcIMSs also
supports remote data query for extraction of specic spatial and
non-spatial feature information, as well as Geocoding, through
which the server assigns an address to a given geographic location
(ArcIMSs Product Overview, 2008).
The ArcIMSs business logic tier components consist of an
Application Server, which handles the load distribution of
incoming requests, the Spatial Servers, used for processing
requests and responses and ArcIMSs Viewers, providing the user
interface and data presentation capabilities. For easy setup of
these services, ESRI provides a Graphical User Interface (GUI) in
the form of ArcIMS Administrator. The ArcIMSs package also
includes ArcIMSs Designer, an easy way to use tool for automatic
generation of ArcIMSs Viewers. The viewers are divided into two
categories: thin DHTML clients, which can be modied using
HTML and JavaScriptTM, and thick Java clients, which can be built
from the ground-up using provided libraries. The clientserver
communication is realized through ArcGISs XML (ArcXML)
requests. ArcXML is also used in constructing image services
including building and styling the map as well as labelling
features (ArcIMSs 9 Architecture and Functionality (2004)An
ESRIs White Paper, 2004).
The ESRIs ArcIMSs is built upon Java technology and thus
requires a Servlet engine as well as a Java Standard Edition
Runtime (JRE). The ArcIMSs version used in the project was 9.1
with Service Pack 2 installed. For running this particular revision,
ESRIs recommends the Tomcat Servlet Engine version 5.0.28
along with the Apache Web Server 2.0.48 and Sun JRE version
1.4.2_06. This is the conguration used for the rst implementa-

tion of the system. Because the Java technology was imposed by

the Internet map server module, it was a logical course of action
to use it for implementing the remaining sub-modules of the Web
GIS.
During the implementation process, all of the Web GIS submodules were named appropriately to their functions in the
system, as shown in the structure of the system, depicted in Fig. 4.
The data adapter module handles the processing of incoming oil
spill data les. All oil-spill related informations is extracted and
stored in a node tree structure in which every branch contains data
for a given hour. This structure is then converted into standard
ESRIs Shapeles (ESRIs Shapele Technical Description, 1998)
through the use of a custom developed algorithm as well as Open
Source JavaDBF libraries (Kumar and Jacob, 2004). Upon completion
of these tasks, the data adapter module passes control to the
IMSHandler module, which is responsible for registering new
Shapeles in the ArcIMSs services. This task is achieved by onthe-y generation of proper ArcXML conguration les and
refreshing the IMS Image Service used for data presentation.
The standard Javascript-based ArcIMSs DHTML Viewer was
used as the remote Client for the Web GIS. The default
conguration of the Viewer, as provided by the ESRIs, allows
for multilayer visualization of geospatial data and includes query
tools, which can be used for extracting detailed information from
selected oil spill layers. The ArcIMSs Author application was used
to design a custom oil spill layer rendering style, depending on the
spatial distribution of its elements. This information, along with
the list of layers available through the Viewer, is stored in the
ArcXML conguration le of the corresponding Image Service.
Other features of the system are provided by the ArcIMSs DHTML
Viewer by default and do not require work from the developer.
These encompass view zooming and panning as well as display of
a miniature overview map in the upper left map corner and a map
legend on the right side of the screen. A sample view of the
resulting web page using the ArcIMSs-based version of
the system is shown in Fig. 5. The end user can see a snapshot
of the oil spill, detected on 13 April 2007 in Aegean Sea, southwest
of Cape Sounio (23.819oE, 37.591oN), after 50 h of spreading;
the information on the bathymetric position of the oil spill
droplets is also provided through a 4-layer visualization scheme
(differentiating droplets with depth greater than 10 m, droplets
with depth between surface and 10 m, droplets that reached land
and droplets lying at the bottom of the sea).
Although the resulting Web GIS served its purpose well, it was
revealed that the technology limitations would not allow for
implementation of some of the planned features. In particular, the
system lacked functionalities such as layer animation (needed for
interactive presentation of oil spill dispersion) and dynamic
changing of layer colour schemes (required to effectively show

1074

M. Kulawiak et al. / Computers & Geosciences 36 (2010) 10691080

Fig. 4. ArcIMSs-based system structure.

Fig. 5. A sample view of resulting web page using ArcIMSs-based version of the system: a snapshot of oil spill, detected on 13 April 2007 in Aegean Sea, is shown
southwest of Cape Sounio (see Fig. 1) after 50 h of spreading; information of bathymetric position of oil spill elements is also provided through a 4-layer visualization
scheme (differentiating elements (I) with depth greater than 10 m, (II) with depth between surface and 10 m, (III) which reached land and (IV) lying at bottom of sea).

spatial distribution of oil spill particles). Examination of the HTML

clients underlying Javascript code led to the conclusion that the
way it updates the screen contents by redrawing the whole map
area makes it unt to serve such types of interactive content. This
issue was not easily foreseen as the ESRIs ArcIMSs lacks detailed
documentation for many of its features. Since changing the code
base for the client meant losing one of the main advantages of
using ArcIMSs, it was a good opportunity to upgrade the entire
system with modern GIS technologies. At rst, the newly released
ESRIs ArcGISs Server 9.2 was considered. However, quick tests
involving management of greater amounts of data revealed
stability and performance issues. Further research revealed that
the amount of potential problems was much greater (e.g. the
functionality of different components tended to be changed, xed
and broken with consecutive service packs) and was generally

attributed to the immaturity of the platform (Nguyen and

Srinivas, 2008). This turn of events provided motivation to build
the new service around open standards. After some research, it
was decided to leave the core functionality on Tomcat and Apache
and serve the resulting Shapeles using Open Geospatial
Consortiums (OGC) Web Map Service (WMS) interface standard,
see (De la Beaujardiere, 2006). Among Open Source map servers,
the two solutions that provided excellent OGC standard compliance, MapServer (McKenna et al., 2009) and GeoServer
(Pumphrey, 2008), offered similar features and performance.
Eventually, the GeoServer option was chosen due to being built
upon Java, which the GUT team was more familiar with. The
GeoServer comes integrated with an Open Source Javascript
library for displaying map data, called the OpenLayers. The library
is very versatile and has been thoroughly tested with GeoServer

M. Kulawiak et al. / Computers & Geosciences 36 (2010) 10691080

1075

Fig. 6. GeoServer-based system structure.

(Pumphrey, 2008), so it was a natural choice for building the

DHTML client. This setup promised to meet all the requirements
set for the Web GIS component, while at the same time it offered
much greater possibilities of customization for the purpose of
implementing features like interactive oil spill animation.
4.2. Implementation of the GeoServer option to oil spill data
The GeoServer is a dynamically developing Open Source J2EE
Servlet-based GIS server, which fully supports the OGC WMS,
Web Coverage Service, Web Feature Service and Web Feature
Service Transactional (WFS-T) protocols (Holmes, 2008). Like the
ArcIMSs, it offers support for a broad range of standard data
sources as well as custom data formats. The latter can be dened
through GeoServers DataStore plugin system, based on the Open
Source GeoTools package (Warmerdam, 2007). The GeoServer
allows the data to be published in the form of images, using one of
the many available output formats or actual data (using WFS), but
unlike ArcIMSs 9.1, it also enables users to update, delete, and
insert geographic information by means of WFS-T, see e.g.
(Vretanos, 2005). An embedded EPSG database allows for realtime data reprojection between hundreds of existing as well as
user-dened projections. This allows data from the GeoServer to
be easily combined with other geographic information (Gui and
Du, 2007). Unlike the ArcIMSs, which stores layer style data in
the services conguration ArcXML le, the presentation of every
layer stored in the GeoServer database can be individually
customized using Styled Layer Descriptor (SLD) les, which can
be dynamically swapped and altered during web queries.
Analogically to commercial solutions, all of the listed features
can be congured from a user-friendly web-based conguration
tool, enabling anyone to quickly share their geospatial data
(Pumphrey, 2008). In order to fully utilize its GIS capabilities, the
GeoServer should be used in conjunction with compatible client
software. While this may be any application supporting open
standards of data exchange, such as Google Earth, UDig, GVSig and
others, the GeoServer is known for its effective integration with
OpenLayers, which is also a dynamically developing Open Source
project.
OpenLayers is a pure object-oriented JavaScript library for
displaying map data in most modern web browsers, with no
server-side dependencies. It implements a constantly developing
JavaScript Application Programming Interface (API) for building
rich web-based geographic applications. Although exhibiting

similarities to the Google Maps and MSN Virtual Earth APIs,

OpenLayers has the advantage of being developed for and by the
Open Source software community, which makes it free for
modication and redistribution. OpenLayers allows the developer
to utilize many readily available and fully customizable components for easy construction of GIS client solutions, which can be
embedded into any Web page. As a framework, OpenLayers is
intended to separate map tools from map data so that all the tools
can operate on all the data sources by means of industry-standard
methods, such as the OGC WMS and WFS protocols (Schmidt,
2008). Unlike the ArcIMSs DHTML Viewer, the default installation
of OpenLayers only provides basic GIS features like zooming and
panning. Advanced GIS features must be added manually by the
developer, using a selection of existing controls as well as their
source code. In practice, however, building a Web GIS client with
OpenLayers is very easy, thanks to the vast library of examples.
This, along with good documentation and support for tiled layer
caches, makes this Open Source product very exible, which
constitutes one of its greatest advantages over the ArcIMSs
DHTML Viewer.
During the process of transition from ESRIs technology to
Open Source solutions, all of the Web GIS modules, except for the
Data Adapter sub-module, were changed accordingly. The Map
Server, the server administration module and the DHTML client
were replaced entirely and a new element was added in the form
of TileCache, as shown in the structure of the GeoServer-based
system, depicted in Fig. 6.
The GeoServerAdmin module is responsible for registering
Shapeles in the GeoServer Database as well as preparation and
setup of a website for viewing the results. The modules
functionality includes addition, removal and modication of
GeoServers feature database entries, appropriately updating
GeoServers XML conguration les, processing oil spill descriptive data (e.g. the number of hours and layer colors) and setting an
initial website environment.
TileCache is an implementation of a Web Mapping ServiceCached (WMS-C), see e.g. (Erle et al., 2007). In the simplest use
case, TileCache requires only write access to a disk and the ability
to run CGI scripts on the server (MetaCarta, 2008). Since the
GeoServer offers tiled layer rendering, TileCache was made part of
the system in order to reduce the server load by caching
background WMS layers. It is particularly useful for large raster
images, which normally require a lot of processing power to
handle (Yang et al., 2005).

1076

M. Kulawiak et al. / Computers & Geosciences 36 (2010) 10691080

Although the process of moving system functionality to a

heavily different software environment was troublesome, it
exhibited specic advantages in respect to the environmental
application under study:

 The ability to serve layers in the form of tiles constituted a




considerable improvement in general system responsiveness

over the previous version. Although the time required to
redraw the entire map was found to be similar for both
solutions, the ArcIMS transmits the map in the form of a single
image while the GeoServer rst sends out a series of tiles.
During the conducted tests the GeoServer would rst transmit
four central tiles, allowing the end user to concentrate on the
area of interest without waiting for other images.
The introduction of the TileCache WMS caching system
substantially reduced system load and improved the speed of
serving background raster layers.
The source code of the OpenLayers library could be freely
modied and extended in order to provide the end user with
specialized tools for layer query and animation as well as data
extraction. An additional benet of using OpenLayers was the
ability to display map layers in the Greek Grid projection while
operating on coordinates in the WGS84 format.
The ability to full WMS GetMap requests using custom
styling for the returned layers allowed for on-the-y creation
of separate maps containing detailed views of selected spill
layers. The importance of this feature is explained in detail in
the next section.

5. Operational application of the Web-based GIS to oil spill

spread prediction
According to the operation of the MARCOAST monitoring and
forecasting service for the Aegean Sea, described in Section 2.1,

data concerning the location and characteristics of the oil spill are
provided from the monitoring module to the forecasting part as
necessary input for the prediction models. The latter produce a
forecast of the oil spill spread for up to 72 h. Following the
developments of the present work, this information is then
conveyed to the Web GIS server for visualization. When registered
users of the service are notied of this fact via email, they may
proceed to log into the Web GIS in order to view the simulation
results. This section illustrates such a scenario using the latest
version of the Web GIS.
Authenticated users are presented with a map of Aegean Sea,
Greece, comprising of several layers. These include raster height
and depth maps as well as vector features showing political
division. Several overlays, depicting (indicatively) protected areas,
national parks, wild life reservoirs and important natural sites, are
also available for selection. A sample view of the main map
containing most of the aforementioned layers in addition to two
oil spills, detected in the Aegean Sea (north of Lesvos Isl. and
southwest of Cape Sounio, see Fig. 1) on 13th of April 2007,
9:27 a.m., is shown in Fig. 7.
The user may utilize the available GIS tools to analyse the oil
spill layers, each of them corresponding to an hourly forecast of
the oil spill spread. In Fig. 8, the user chose to concentrate on the
oil spill located southwest of Cape Sounio (it poses a potential
threat to nearby protected natural sites and densely populated
areas) and may proceed to analyse the evolution of the oil spill by
launching an animation of its spread throughout the simulated
period, as shown in Fig. 9.
Fig. 10 depicts the drift of the oil spill throughout the rst 12 h
of the simulation period, as shown by the layer animation tool.
The user can follow the drift of the oil spill, which moves to the
southeast.
The complete simulated path of the oil spill may be observed
by either turning on all available layers or narrowing the
selection to layers of the same colour, which are placed at

Fig. 7. Sample view of main map (Aegean Sea, Greece) that authenticated users of the system are presented, showing two oil spills detected north of Lesvos Isl. and
southwest of Cape Sounio on 13th of April 2007, 9:27 a.m. Map also contains several overlays depicting protected areas, national parks, wild life reservoirs and important
natural sites, which may be potentially endangered by spread of oil spills.

M. Kulawiak et al. / Computers & Geosciences 36 (2010) 10691080

1077

Fig. 8. Utilization of ZoomBox control to concentrate on oil spill located southwest of Cape Sounio: one of the available GIS tools for analysis of available oil spill layers,
which represent consecutive hours of spread forecast.

Fig. 9. Spill simulation control on left side of screen may be used for easy analysis of forecast of oil spill spread. Clicking on control launches an animation of oil spill spread
throughout the 50 h period of simulated spread.

regular time intervals. A comparison of these two methods is

shown in Fig. 11.
Because an oil spill is subject to phenomena such as evaporation,
emulsication, beaching and sedimentation, its properties will

change with time. This may lead to pollution of different elements

of the environment, for example when oil settles on the sea oor due
to sedimentation. Since the above phenomena are simulated by the
operational model, the user may wish to examine particular spill

1078

M. Kulawiak et al. / Computers & Geosciences 36 (2010) 10691080

Fig. 10. Forecast of oil spill drift throughout rst 12 h of simulation, as shown by layer animation tool.

Fig. 11. Two ways of observing complete simulated path of oil spill spread: method 1 involves turning on all available layers (left image); method 2 allows to narrow
selection to layers of the same color, which are placed on layer list at regular time intervals (right image).

layers in more detail. Such information is provided by additional

windows, which appear when a user clicks on the Details button
next to the name of a layer in the list on the right side of the map
screen. These windows consist of an independent map, containing
data from a single-selected oil spill, drawn with depth-dependent
coloring and overlaid on thematic background layers, as well as oil
volume statistics. Fig. 12 shows a detailed view of the oil spill

detected on 13th of April 2007 north of Lesvos Isl. at its 43rd hour of
simulated evolution.
As it may be seen in gure, by 3:27 a.m. (local time) on
15.04.2007, it is estimated that approximately 23% of the oil spill
will have shifted over 10 m below the surface of the sea, while by
that time no elements of the spill will have reached either land or
the bottom of the sea.

M. Kulawiak et al. / Computers & Geosciences 36 (2010) 10691080

1079

Fig. 12. A detailed view of oil spill detected on 13th of April 2007, north of Lesvos Isl. at its 43rd hour of simulation, available in an independent map window. Oil spill layer
is presented in depth-dependent coloring and accompanied by volume statistics. The latter show that, by 3:27 a.m. (local time) on 15.04.2007, approximately 23% of the oil
spill would have shifted over 10 m below the surface of sea, but no oil particles would reach either land or bottom of the sea.

6. Summary and conclusions

The main goal of this work was to implement a Web-based
Geographic Information System for an existing operational oil
spill monitoring and forecasting service, developed in the framework of the MARCOAST project. The oil spill service for the Aegean
Sea is based on the routine acquisition and analysis of SAR images
and the spreading of the detected oil spills by means of the
complex POSEIDON operational tool for the prediction of oating
pollutant transport, providing hourly forecasts of the spill spread
for up to 72 h after the time of event. The output information of
the aforementioned MARCOAST service for two oil spill spreading
events in two different areas of the Aegean Sea was provided as
input in order to test the capabilities for visualization and
mapping and the general performance of the Web-based GIS
under development.
During the course of development of the presented system, a
lot of working time was spent on various web mapping
technologies, which brought up some interesting points regarding
the evolution of Web GIS mapping solutions and the changing
position of Open Source software versus its commercial counterparts. In terms of application development time, commercial
solutions like the ArcIMSs offer automated tools, which allow for
quick and easy creation, styling and setup of a functional Web
service for serving thematic geospatial data. A similar Website
created with Open Source tools would require weeks of development. Available technical support for both ArcIMSs and GeoServer is generally comparable. Their setup and basic operations are
equally well described, while more advanced functionalities are
similarly difcult to use because of sparse documentation. The
time required for generation of the whole visible map area was
found to be similar for both solutions; however overall system
responsiveness of the GeoServer option was better due to its tiled
layer rendering. Finally, in terms of exibility, the combination of
GeoServer and OpenLayers allowed for implementing all of the

designed functionalities for the Web GIS, some of which could not
be achieved with the ArcIMSs. Since the end of the discussed
project, all of the described technologies have progressed. The
ESRIs ArcIMSs has been superseded with the ArcGISs Server 9.3,
which, in addition to support for tiled map caches, offers better
speed and improved stability over its predecessors. At the same
time, GeoServer 2.0 and OpenLayers 2.8 have been vastly
enhanced in terms of functionality and ease of use. Although
the general conclusions regarding comparison of commercial and
free technologies still stand, it has now become much easier to
build an advanced Web Service based on Open Source solutions.
Concerning the presented environmental application, the
interactive visualization of marine pollution monitoring and
forecasting data has been achieved by the remote presentation
via a dedicated Web GIS, which allows authenticated end users to
view the simulation results in the form of thematic layers set in a
geographical context and overlaid on background data. The initial
ArcIMSs version of the system presented considerable improvements over the MARCOAST Web-based dissemination service by
presenting oil spill forecasts through a fully functional GIS.
The nal GeoServer version of the Web GIS offered further
improvements in the form of better responsiveness, enhanced
layer styling, display of layer animations and an enhanced user
interface. The added value of presenting comprehensive and
synthetic, both temporal and spatial, environmental information
through an interactive, easy to navigate graphical interface
renders the developed Web GIS easily accessible even to
inexperienced users. Elements of the new system were adapted
for presentation of oil spill spread forecasts in the Baltic Sea and
have been well received by representatives of the Polish Maritime
Institute in Gdansk (Kulawiak et al., 2010). In this context it could
be said that, although the presented Web GIS application requires
some nalization steps, there are no major obstacles in the way of
its further development to an enhanced pollution awareness and
emergency management tool.

1080

M. Kulawiak et al. / Computers & Geosciences 36 (2010) 10691080

Acknowledgements
The oil spill dispersion data used in the study were produced in
the framework of the MARCOAST project, funded by ESA and
coordinated by Thales Alenia Space, contract number: 136391.
The new concept for visualization of the oil spill simulation
results was developed in the framework of Joint Polish-Greek R&T
Programme 20052007 Development of a monitoring and
mapping system for marine pollution via a Web-based Geographical Information System. HCMR was nancially supported by
the Greek General Secretariat for Research and Technology.

References
Al-Sabhan, W., Mulligan, M., Blackburn, G.A., 2003. A real-time hydrological model
for ood prediction using GIS and the WWW. Computer. Environment and
Urban Systems 27, 932.
ArcIMSs 9 Architecture and FunctionalityAn ESRIs White Paper, 2004. EDN:
ESRIs Developer Network, Environmental Systems Research Institute, Inc.,
/http://www.esri.com/library/whitepapers/pdfs/arcims9-architecture.pdfS,
[Accessed April 29, 2009].
ArcIMSs Product Overview, 2008. Environmental Systems Research Institute, Inc.,
/http://www.esri.com/library/brochures/pdfs/arcims_bro.pdfS,
[Accessed
October 22, 2009].
Blumberg, A.F., Mellor, G.L., 1987. A description of a three-dimensional coastal
ocean circulation model. In: Heaps, N.S. (Ed.), Three-Dimensional Coastal
Ocean Models, Coastal and Estuarine Science, No. 4. American Geophysical
Union, Washington, DC, pp. 116.
Chang, Y.-S., Park, H.-D., 2004. Development of a web-based geographic
information system for the management of borehole and geological data.
Computers & Geosciences 30, 887897.
De la Beaujardiere, J. (Ed.), 2006. OpenGIS Web Map Service (WMS) Implementation Specication ver. 1.3.0. Open Geospatial Consortium, Inc. Accessed
October 22, 2009.
Doyle, S., Dodge, M., Smith, A., 1998. The potential of Web-based mapping and
virtual reality technologies for modelling urban environments. Computer,
Environment and Urban Systems 22 (2), 137155.
Erle, S., Bitner, D., Spencer, P., 2007. WMS Tile Caching. The Open Source
Geospatial Foundation. /http://wiki.osgeo.org/wiki/WMS_Tile_CachingS,
[Accessed October 22, 2009].
ESRIs Shapele Technical Description, 1998. EDN: ESRIs Developer
Network. Environmental Systems Research Institute, Inc. Accessed October
22, 2009.
Gui, Z., Du, J., 2007. Mapping distributed trafc information based on
GeoRSS&WMS, Proceedings of the International Conference on Convergence
Information Technology. IEEE Computer Society, Gyeongju-si, Gyeongbuk,
Korea, pp. 16341637.
Hamre, T., Krasemann, H., Groom, S., Dunne, D., Breitbach, G., Hackett, B., Srensen,
K., Sandven, S., 2009. Interoperable web GIS services for marine pollution
monitoring and forecasting. Journal of Coastal Conservation 13 (1), 113.
Holmes, C. (Ed.), 2008. GeoServer Features. The GeoServer Project., /http://
geoserver.org/display/GEOS/FeaturesS, [Accessed October 22, 2009].
Kontoes, C.C., Sykioti, O., Paronis, D., Harisi, A., 2005. Evaluating the performance of
the space-borne SAR sensor systems for oil spill detection and sea monitoring
over the south-eastern Mediterranean Sea. International Journal of Remote
Sensing 26 (18), 40294044.
Kraak, M.-J., 2004. The role of the map in a Web-GIS environment. Journal of
Geographical Systems 6, 8393.
Kruger, A., Khandelwal, S.G., Bradley, A., 2007. AHPSVER: A web-based system for
hydrologic forecast verication. Computers & Geosciences 33, 739748.
Kulawiak, M., Chybicki, A., Moszynski, M., 2010. Web-based GIS as a tool for
supporting marine research. Marine Geodesy, in press.
Kumar K.A., Jacob, A.K., 2004. JavaDBF Documentation. Sarovar.org Open Source
project portal. /http://sarovar.org/docman/index.php?group_id=32&selected_
doc_group_id=29&language_id=1S, [Accessed 28 April, 2009].

McKenna, J., Fawcett, D., Butler, H., 2009. An Introduction to MapServer. The
MapServer Project. /http://mapserver.org/introduction.htmlS, [Accessed 28
April, 2009].
MetaCarta, 2008. TileCacheWeb Map Tile Caching. MetaCarta Labs. /http://
tilecache.orgS, [Accessed October 22, 2009].
Nittis, K., Perivoliotis, L., Korres, G., Tziavos, C., Thanos, I., 2006. Operational
monitoring and forecasting for marine environmental applications in the
Aegean Sea. Environmental Modelling & Software 21, 243257.
Nguyen, C., Srinivas, J., 2008. A Reusable Mapping Platform Based on ArcGISs
Server 9.2. Environmental Systems Research Institute, Inc. accessed 04
September, 2009.
Papadopoulos, A., Kallos, G., Katsafados, P., Nickovic, S., 2002. The weather
forecasting system for Poseidon: An overview. The Global Atmosphere and
Ocean Systems 8 (23), 219237.
Pavlakis, P., Tarchi, D., Sieber, A.J., Ferraro, G., Vincent, G., 2001. On the monitoring
of illicit vessel dischargesa reconnaissance study in the Mediterranean Sea.
EC DGJoint Research Council, EC DGEnvironment, European Commission
Report EUR 19906 EN, 21pp. /http://ec.europa.eu/environment/civil/marin/
reports_publications/jrc_illicit_study.pdfS, [Accessed October 22, 2009].
Petihakis, G.I., Triantafyllou, G.N., Koutitas, C.G., 2001. Prediction and prevention of
oil contamination and monitoring of the benthic structure and related sheries
in connection with the pollution impact. Systems Analysis Modelling
Simulation 41 (1), 169197.
Pollani, A., 2005. Modeling oil spills in areas of complex topography. In: Simos, T.,
Maroulis, G. (Eds.), Advances in Computational Methods in Sciences and
Engineering 2005, vol. 4 A. Brill Academic Publishers, Leiden, The Netherlands,
pp. 866871.
Pollani, A., Triantafyllou, G., Petihakis, G., Nittis, K., Dounas, K., Koutitas, C., 2001.
The POSEIDON operational tool for the prediction of oating pollutant
transport. Marine Pollution Bulletin 43 (712), 270278.
Prospathopoulos, A., Kulawiak, M., Kioroglou, S., Perivoliotis, L., uba, M.,
Grammatikaki, M., Stepnowski, A., 2008. Integrating a Web-based GIS to a
marine pollution monitoring and forecasting system. In: Iona, S., Maillard, C.,
Tosello, V. (Eds.), Book of Abstracts, International Conference on Marine Data
and Information Systems (IMDIS-2008), 31 March-2 April, pp. 156.
Pumphrey, M. (Ed.), 2008. What is GeoServer. The GeoServer Project. /http://
geoserver.org/display/GEOS/What + is + GeoserverS, [Accessed October 22,
2009].
Rao, M., Fan, G., Thomas, J., Cherian, G., Chudiwale, V., Awawdeh, M., 2007. A webbased GIS decision support system for managing and planning USDAs
Conservation Reserve Program (CRP). Environmental Modelling & Software
22, 12701280.
Schmidt, C., 2008. OpenLayers: Free Maps for the Web. /http://www.openlayers.
orgS, [Accessed October 22, 2009].
Soukissian, T., Chronis, G., Poseidon Group, Ballas, D., Vlahos, D., Nittis, K.,
Diamanti, Ch., Perivoliotis, L., Papageorgiou, E., Barbetseas, S., Mallios, A., 2000.
Poseidon: a marine environmental monitoring, forecasting and information
system for the Greek Seas. Mediterranean Marine Science 1, 7178.
Su, Y., Slottow, J., Mozes, A., 2000. Distributing proprietary geographic data on the
world wide web-UCLA GIS database and map server. Computers & Geosciences
26, 741749.
Sugumaran, R., Meyer, J.C., Davis, J., 2004. A Web-based environmental decision
support system (WEDSS) for environmental planning and watershed management. Journal of Geographical Systems 6, 307322.
Tsou, M.-H., 2004. Integrating web-based GIS and image processing tools for
environmental monitoring and natural resource management. Journal of
Geographical Systems 6, 155174.
, Hamre, T., 2007. Design and implementation of a distributed GIS portal
Tuama, EO
for oil spill and harmful algal bloom monitoring in the marine environment.
Marine Geodesy 30, 145168.
The WAMDI Group, 1988. The WAM modelA third generation ocean wave
prediction model. Journal of Physical Oceanography 18 (12), 17751810.
Vretanos, P.A. (Ed.), 2005. OpenGIS Web Feature Service (WFS) Implementation
Specication ver. 1.1. Open Geospatial Consortium Inc. Accessed October 22,
2009.
Warmerdam, F., 2007. GeoTools InfoSheet. Open Source Geospatial Foundation.
/http://www.osgeo.org/content/projects/geotools.htmlS,
[accessed
03
September, 2009].
Yang, C.P., Wong, D.W., Yang, R., Kafatos, M., Li, Q., 2005. Performance-improving
techniques in web-based GIS. International Journal of Geographical Information Science 19 (3), 319342.