lable at ScienceDirect

Environmental Modelling & Software 77 (2016) 50e62

Contents lists avai

Environmental Modelling & Software

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

An innovative system for sharing, integration and visualization ofheterogeneous 4D-information

R. Paciello a, b, *, I. Coviello a, b, P. Bitonto c, A. Donvito c, C. Filizzola b, N. Genzano a, d,M. Lisi a, d, N. Pergola a, b, G. Sileo a, V. Tramutoli a, b, e

a University of Basilicata, School of Engineering, Potenza, 85100, Italyb National Research Council (CNR), Institute of Methodologies for Environmental Analysis (IMAA), Tito Scalo (PZ), 85050, Italyc Digimat S.r.l., Matera, 75100, Italyd GeoSpazio Italia S.r.l., Potenza, 85100, Italye International Space Science Institute (ISSI), Bern, 3012, Switzerland

a r t i c l e i n f o

Article history:Received 29 September 2014Received in revised form5 November 2015Accepted 20 November 2015Available online xxx

Keywords:Satellite observationsHeterogeneous spatiotemporal dataintegrationAsynchronous data visualizationTime-varying datasetsEarthquake precursors4DEOS

* Corresponding author. National Research Counciologies for Environmental Analysis (IMAA), Tito Scalo

E-mail address: rossana.paciello@imaa.cnr.it (R. Pa

http://dx.doi.org/10.1016/j.envsoft.2015.11.0111364-8152/© 2015 Elsevier Ltd. All rights reserved.

a b s t r a c t

Treating the time dimension (i.e., the 4th dimension) of geospatial parameters, likewise the other di-mensions is important for some applications to better understand the possible space-time relationsamong the physical parameters underlying the dynamics of complex phenomena. In this paper wepresent the potential of an innovative solution, named 4 Dimensions Environmental ObServation(4DEOS), which is a platform able to handle the 4th dimension in an asynchronous way (i.e., to visualizemore signals, holding some of them fixed in the time domain while moving others over time). 4DEOS isbased on a Client-Broker-Server architecture for the easy integration and visualization of heterogeneous,asynchronous, geospatial products. The prototype system was evaluated in the case of earthquake pre-diction studies where the asynchronous visualization of independent observations could allow a timelyidentification of the spatial correlations appearing at different time lags, which could be missed usingother existing 4D Geographic Information System software.

© 2015 Elsevier Ltd. All rights reserved.

Software availability

Software name: 4DEOS (4 Dimensions Environmental ObServation)Developers: Rossana Paciello, Irina Coviello and Paolo BitontoYear first official release: 2012Hardware requirements: standard PCSystem requirements: Microsoft Windows (XP or later), Java

(version 1.6 or higher)Program language: JavaProgram size: 420 MBAvailability: “Downloads” area in http://www.pre-earthquakes.org

(user registration is required)License: Free under a GNU General Public License (www.gnu.org)

agreementDocumentation for users: Manual

l (CNR), Institute of Method-(PZ), 85050, Italy.ciello).

1. Introduction

The analysis of complex natural phenomena often requires theintegration of heterogeneous time-dependent datasets in order toinvestigate the possible correlations among the underlying physicaland chemical parameters. Depending on the spatial and temporaldynamics of each parameter, such datasets could have a differentspatial resolution and could be asynchronous (i.e., not recordedtogether, at same time) and/or asynchronously visualized for spe-cific applications.

Ascertaining the occurrence of spatiotemporal correlationsamong the recorded signals, also at different time lags, is crucial formany applications. For instance, in the case of flood forecasting, ananomalous variation of the soil moisture measured by meteoro-logical satellites (e.g., Lacava et al., 2010) could be compared withthe precipitation forecasts for the following days to dynamicallyupdate the flood risk estimates. Also in the case of earthquakeevents (e.g., Bonfanti et al., 2012) and volcanic eruptions (e.g.,Marchese et al., 2014), anomalous signals coming from a multi-parametric observation system and appearing at different times,

R. Paciello et al. / Environmental Modelling & Software 77 (2016) 50e62 51

could be better compared each other using tools able to dynami-cally and asynchronously visualize multi-parametric geospatialdatasets. This suggests that the temporal dimension should havethe same importance as the other ones.

The explicit representation and analysis of spatiotemporal datawere first theorized by Langran and Chrisman (1988). Since thebeginning of the 1990s, several efforts have been made to buildgeographic databases and platforms (Abraham and Roddick, 1999;Langran, 1993; Peuquet and Duan, 1995; Rasinmaki, 2003; Galton,2004; Pelekis et al., 2004; Zhou et al., 2004; Le, 2005; Peuquet,2005; Zhu et al., 2008; Fan et al., 2010) specifically devoted tospace-time data management and analysis. Different approacheswere proposed and various data models and platforms weredeveloped, also supported by technological advancements(Maceachren et al., 1999; Tryfona and Jensen, 1999; Shi and Zhang,2000; Wang et al., 2005; Lee et al., 2006; Shaw et al., 2008; Shawand Yu, 2009; Vivar and Ferreira, 2009; Pultar et al., 2010;Sadahiro, 2002; Sugam et al., 2013; Gebbert and Pebesma, 2014).

At present spatiotemporal correlation analysis and time-varyingsignal visualization are typically performed using geospatial datamanagement systems like Geographic Information System (GIS)and their derivatives that allow users to overlay spatial data and tomanage time as an attribute of the spatial reference (Pelekis et al.,2004). This kind of architecture, however, does not incorporate thetemporal indexing into the GIS itself (Peuquet and Duan, 1995;Peuquet, 2001; Galton, 2004) and the representation of asynchro-nous datasets over time does not take into account the full di-mensions of a dataset and its continuous or discrete variations.Consequently, customized systems have to be designed anddeveloped in each different application to properly manage data inthe space-time domain (Le, 2012). Present geospatial data man-agement systems permit users to visualize spatial, time-dependentsignals thanks to specific functions, but the time variations of suchsignals are visualized overlapping one another, with the conse-quence that it is impossible to move a specific signal over time andvisualize its time variations while keeping the others fixed. The 4Dimensions Environmental ObServation platform (4DEOS), aClient-Broker-Server (CBS)1 platform, here presented is able tovisualize together several asynchronous spatiotemporal data/products provided by servers and queryable by authorized users.The innovative aspect of 4DEOS is quite evident, giving the platformthe possibility of visualizing spatial, time-dependent signals in asimultaneous or asynchronous mode so that the study of theirtemporal evolution may be more easily carried out. Users can moveone or more signals while holding others fixed in the time domain.

In the following paragraphs we describe the detailed 4DEOSarchitecture and its utilization to study a particular application: thedetection of earthquake precursor signals, developed in theframework of the EU-FP7 Project PRE-EARTHQUAKES (ProcessingRussian and European EARTH Observations for earthQUAKE pre-cursors Studies, http://www.pre-earthquakes.org).

2. The 4DEOS architecture

The 4DEOS platform was designed and developed with the aimof offering clients a single entry-point to visualize heterogeneousdata (e.g., maps, vertical profiles, punctual time series related todifferent observation times and/or geographic areas) shared bydifferent data providers. 4DEOS does not use a standard client/

1 The Client-Broker-Server pattern architecture allows clients to access remoteservers through the broker, a “middle” component, whose responsibility is thetransmission of requests from clients to servers, as well as the transmission ofresponses back to the client.

server architecture, where a client, needing a service from aparticular server, sends a request to an appropriate server and theserver performs the requested service, returning results to theclient. This client/server architecture works fine only for smallsystems (i.e., when clients interact with a small set of servers -Adler, 1995), while, in growing and evolving systems like 4DEOS,individual clients need to be constantly updated to take into ac-count the addition of new servers and services (Adebayo et al.,1997).

4DEOS is based on a service-oriented CBS architecture intro-ducing a middle component between clients and servers, calledbroker, which receives requests from clients, identifies appropriateremote servers, forwards requests to servers and transmits resultsto clients. It maintains centralized information (e.g., data availablein each remote server, authorized clients as well as data that can berequested by each of them) so that clients need to know nothingabout the existing servers, just how to interact with the broker(Fig. 1). Each 4DEOS server node has the task of producing and (ifnecessary) converting with specific converter tools data/productsfrom a particular data provider format to a standard one (e.g., shapefile format). This task is typically performed by the brokercomponent, whereas in the 4DEOS architecture this role is assignedto the servers, because the data sharing process is made in a contextwhere data owners (server nodes) are generally reluctant to anactual transfer of files, especially when their data have an addedvalue or are very expensive to acquire and maintain. In the 4DEOSarchitecture owners expose their data/products via web services,following Open Geospatial Consortium (OGC) standards (http://www.opengeospatial.org; Open Geospatial Consortium, 2005,2006, 2008), which can be queried only by the broker compo-nent. In particular, each server node is configuredwith a set of opensource software:

� Converter tool, different and specific for each single node. It iswritten in Java and performs the conversion of data/productsfrom the provider characteristic format to a standardized vectorformat (i.e., shape file);

� Ingestor tool, based on GeoBatch Java software (http://geobatch.geo-solutions.it). It stores shape files in a PostgreSQL databasewith POSTGIS extension (http://postgis.refractions.net/) andpublishes data/products as OGC WMS (Web Map Service) ser-vices in GeoServer (http://geoserver.org/).

Besides passing the requests from clients to servers, the brokercomponent manages the security policy that, in order to supportand encourage data sharing, was deliberately based on the conceptthat “the more data you share, the more data you obtain”. Thisprinciple, here abbreviated MOMO (More Offer More Obtain), wasintroduced to provide data access according to each provider'scontribution; for example, if a data provider shares data relating toa specific geographic area, he will be able to view only the datashared by other providers that refer to the same geographic area.This concept is implemented in the broker by using 52� North WebSecurity Service (WSS, http://52north.org/) to define multiplelevels of data access authorizations. Thanks to XML configurationfiles, the broker retains information about i) the data/productsshared by each server node; ii) the clients enabled to access the4DEOS platform and iii) the data/product use restrictions con-cerning each client (for instance geographic areas).

It is possible to add a new server node to the 4DEOS system atany time provided that the system administrator deems the data/products useful for a specific application. In this case, the brokeradministrator adds to the list of available data the WMS servicesprovided by the new server node and, each 4DEOS Client canvisualize the new data/products at the next access (if enabled). In

Fig. 1. 4DEOS is based on a three-tier distributed architecture: Client-Broker-Server. The server side consists of nodes which are provided with a Converter tool, different andspecific for each node, performing the conversion of data/products from the provider characteristic format to a standardized vector format (i.e., shape file), and with an Ingestor toolstoring files in a database and exposing them as Web Map Service (WMS). The broker side consists of one node that receives requests from clients, identifies appropriate remoteservers, forwards requests to servers and transmits results to clients; furthermore it manages security and data access privileges. The client side refers to a tool that performsvisualization of geospatial products shared by different data providers.

R. Paciello et al. / Environmental Modelling & Software 77 (2016) 50e6252

fact, each time a 4DEOS Client logs on to the system it requests theupdated list of data available to the broker.

The 4DEOS Client is the innovative component of this archi-tecture. A detailed description of how it works is given in thefollowing paragraphs.

3. The 4DEOS client

The 4DEOS Client is a new tool for both the timely access and theeasy visualization and integration of geospatial products havingdifferent space-time dimensions, such as:

� punctual data such as data acquired in a geographical location(x,y) representing the variability of a considered signal at aspecific depth/altitude (z) and its evolution over time (t), like themeasures from meteorological weather stations;

� linear profiles such as data acquired along a linear pattern (e.g.,sounding sensors/satellite tracks, geophysical surveys), repre-senting the evolution over time (t) of a considered signal along aline, like the electrical resistivity profiles or tomographies;

� areal data such as data acquired/retrieved by satellite sensorsand/or spatial data obtained after processing ground data (x,y)or vertical sections along a transect on the ground (x,z) or (y,z),like the land cover maps.

This paper focuses on the 4DEOS Client as a tool to allow usersinvestigations of asynchronous observations over time. The otherproduct capabilities are here intentionally omitted. The softwarehas an innovative feature enabling users to move over time and to

visualize signals in simultaneous or asynchronous way, with thepossibility, in this latter case, of freezing time for some signals tofocus on others to study their temporal evolution better. Such4DEOS Client functionality is achieved through a combined use ofgraphic objects (i.e., sliding time bars) available for any data/product to be investigated, which allow the user to disable the“time management” so to freeze in the time domain some signalsthat continue to be displayed on the map. This feature is not sup-ported in any other existing GIS software in which disabling thetime management feature for certain data/products implies thatthey are no longer displayed on the map.

The investigation of data/products carried out by disabling thetime management feature is an operation simplified by the possi-bility of displaying all signals in a double mode (i.e., in a single or ina dedicated window). This characteristic offered by the 4DEOSClient is not available in any other GIS software and will bedescribed in detail in Sections 3.1 and 3.2.

3.1. Implementation details

The 4DEOS Client is a Java desktop Rich Client Application builton Eclipse “Rich Client Platform” (RCP, http://www.eclipse.org) as acustomized extension of the uDig (http://udig.refractions.net) GISsoftware core. Both Eclipse RCP and uDig are open source andwidely popular tools which enable customization. They are alsocomprehensive, mature, and maintained. They already includemany features necessary for the purpose of this work and, conse-quently, represented our final choice for the 4DEOS Clientdevelopment.

R. Paciello et al. / Environmental Modelling & Software 77 (2016) 50e62 53

Like the uDig software, the 4DEOS Client is built around theconcept of the plug-ins2 at the base of Eclipse RCP. A set of newdedicated plug-ins were developed to achieve all planned specifi-cations. Three plug-ins have been developed in particular: one forthe time management, another for the graphical object construc-tion serving the same purpose, and the last one for searchingavailable data/products. Moreover, time management was alsoadded to the plug-in that supportsWMS requests (already includedin the uDig core software).

The 4DEOS Client was fully developed with open software and iswidely based on libraries of proven reliability that support stan-dards able to ensure interoperability between components. It uses:

i) Java Topology Suite (JTS, http://www.vividsolutions.com/jts/JTSHome.htm), which provides an implementation of theSimple Feature for Structured Query Language (SQL);

ii) GeoAPI (http://www.geoapi.org/), a common Java interfacefor geospatial concepts based on OGC standards;

iii) GeoTools (http://www.geotools.org/), a Lesser General PublicLicense (LGPL) library provided by the Open Source Geo-spatial Foundation (OSGF, http://www.osgeo.org/) for com-mon GIS functionality.

The 4DEOS Client exploits three important extensions of thedefault Java Runtime Environment (JRE): Java Advanced Imaging(JAI, http://www.oracle.com/technetwork/java/javase/tech/jai-142803.html), about the image processing tasks; Java ImageIOthat provides raw raster format and ImageIO-EXT (https://java.net/projects/imageio-ext), to manage additional geospatial rasterformats.

3.2. Functionality details

Awizard (i.e., a graphic user interface consisting of a sequence ofdialog boxes) guides the user through a series of well-defined stepsfrom the selection of a specific place and time interval to thevisualization of the list of available data/products returned by thebroker. Selected products can be then displayed on the mapchoosing between two distinct data visualization modes:

� all data/products are visualized overlapped in a single window(Fig. 2);

� data/products are displayed in a dedicated window (i.e., onewindow for each product, Fig. 3).

The second visualization mode should be preferred when thenumber of areal data (maps) to be superimposed is greater thanthree that is when the large number of data does not allow the userto appropriately appreciate data content.

Each graphic object (the sliding time bar indicated by the redcircles with number 1 in Figs. 2e3) allows the user to follow thetemporal evolution of selected signals. In this way, each maprelated to each single signal presents a static image that can beindividuallymoved through time (in forward or reverse) by clickingon its time bar as this release does not support the animation ofmaps.

The second way to perform signal visualization in the temporaldomain is another innovative feature of the 4DEOS Client, whichhas not been implemented in any traditional GIS software so far.Thanks to this innovative solution, the user can perform a visualanalysis in a completely new manner (i.e., fixing the view of

2 A plug-in is a software component that adds a specific feature to an existingsoftware application.

products containing interesting signals while moving other signalsover time). This way it is possible to make a visual comparisonamong variables taken at different time points within establishedspace-time constraints.

In addition to that, a “global time bar” (see green circle withnumber 2 in Figs. 2e3) allows users to visualize the temporalevolution of all signals simultaneously. Date and time related to theproducts shown in the maps are always visualized over the (localand global) bars by the current time label.

Figs. 2 and 3 give an example of data visualization with the4DEOS Client over a long period of time (about fromAugust 2007 toAugust 2014). The pictures show the integration of fourparameters:

1. Plasma frequency (GRACE/CHAMP);2. GPS TEC variability index punctual time series;3. Electrons density (Formosat-3/COSMIC);4. TIR map (MSG/SEVIRI satellite).

Parameters 1 and 4 are “fixed” at two different time instantswhere it is possible to note some interesting signals, while pa-rameters 2 and 3 are not fixed and, consequently, move when theglobal time bar moves. Note that to fix a signal, the graphic objectplaced in its local time bar (the violet circles in Figs. 2e3) has to beunchecked and that if a map is “fixed” it will not change althoughone moves along the global time bar.

Differently from other traditional software (i.e., ArcGIS 10, QGIS2.0.1), when the user moves along the time axis it is not required toselect a fixed time stamp because the bar will move automaticallyto the next available date and all (not fixed) signals will be auto-matically updated at their last available date within a time intervalthat can be configured for each signal. It means that for eachselected signal there will always be at least one recurrence dis-played on the map.

Choosing the right fixed time stamp is important to work easilywith a software that offers timemanagement capabilities using thisapproach; it is another peculiar aspect of the 4DEOS Client. With atraditional GIS software, if the user chooses a “confined” fixed timestamp (such as few minutes or seconds), he will have to clickseveral times along the timeline before having the data/productsavailable at a wide temporal interval (i.e., at a distance of weeks ormonths) displayed. Conversely, if the user chooses a “wide” fixedtime stamp, a traditional GIS software will cluster a lot of data/products referring to a close temporal distance into a single view,with the consequence that too many data/products are displayedon the map. The 4DEOS Client solves this problem since it does notrequire a fixed time stamp for the sliding time bar, and it allows theuser to automatically visualize the temporal evolution moving tothe next available date.

The 4DEOS Client graphic interface was designed to be user-friendly so to facilitate visual cross-correlations among signalswith different temporal dynamics. In our approach, the time modelimplemented by the 4DEOS Client follows the discrete model(Frank, 1998), including time instants that are unevenly spaced andtime intervals that can overlap or contain time instants.

The whole time management was implemented involving allthree components of the 4DEOS platform. On the server side, data/products (which are available at different nodes) are appropriatelyconfigured to support the time dimension (by enabling the “TIMEDIMENSION00 on the GeoServer and choosing the field date for eachlayer). On the broker side, a specific web service that, at each 4DEOSClient access, requests to all server nodes a date list of their shareddata/products was developed. On the client side, new plug-ins (asrequired by Eclipse RCP) were developed in order to:

Fig. 2. 4DEOS Client view mode: visualization of several signals overlapped in a single view. Close to the map there are as many local time bars as the number of signals (1). In (2) there is a global time bar allowing the user to follow thetemporal evolution of all signals. The violet circles indicate the graphic object that fix a signal (if it is unchecked the map is “fixed” and it will not be change although one moves along the global time bar). (For interpretation of thereferences to colour in this figure legend, the reader is referred to the web version of this article.)

R.Pacielloet

al./Environm

entalModelling

&Softw

are77

(2016)50

e62

54

Fig. 3. 4DEOS Client view mode: visualization of each signal in a dedicated window. Below each map a local time bar allows the user to follow the temporal evolution of the signal (1). In (2) a global time bar allows the user to follow thetemporal evolution of all signals together. The violet circles indicate the graphic object that fix a signal (if it is unchecked the map is “fixed” and it will not be change although one moves along the global time bar). (For interpretation ofthe references to colour in this figure legend, the reader is referred to the web version of this article.)

R.Pacielloet

al./Environm

entalModelling

&Softw

are77

(2016)50

e62

55

R. Paciello et al. / Environmental Modelling & Software 77 (2016) 50e6256

� include the TIME attribute into WMS requests to display on themap data/products referring to a specific time interval;

� create the sliding time bars: starting from the date list (relatedto all available data/products) returned by the broker (by meansof the web service), the software checks those falling within theperiod of time chosen by the user in the search phase and in-cludes them into the global time bar; only the dates of data/products will be used to build each single local time bar.

4. 4DEOS platform application: a test case for the study ofearthquake precursors

The 4DEOS system's potential was evaluated in the case ofearthquake prediction studies where the possibility of integratingobservations of different parameters is expected to stronglyimprove the reliability of short-term earthquake forecasts at pre-sent mostly based on the analysis of foreshocks sequences.

Pre-seismic anomalies of different parameters (such as seis-micity, electric and magnetic fields, gas emissions, surface de-formations, temperature changes, etc.) were reported to occur atdifferent time lags before an earthquake occurrence. One of the firststudies on this topic, Scholz et al. (1973) made a distinction be-tween the short-term precursors which precede an earthquakefrom a few hours to days and the long-term precursors, whichprecede an earthquake from months to years. In a more recentpaper supported by increased scientific literature, Cicerone et al.(2009, and references therein) summarized the results obtainedin more than thirty years of research and reported examples ofearthquake precursors occurring well in advance of the time of anearthquake. This is the case, for instance, of the variations ofionospheric ULF3 and VLF4 emissions (measured by 19 Intercosmossatellites) detected from 8 h before up to 3 h after an earthquakeoccurrence, or the surface deformations measured from months todays before an earthquake.

An asynchronous mode of comparison is consequently neededto fully appreciate the advantage of integrating independent pa-rameters. Having such a specific capability the 4DEOS platformwasused as a common integration and visualization platform in theframework of the PRE-EARTHQUAKES project, the first Europeanproject devoted to investigating the potential of a multi-parametricapproach to short-term earthquake forecast. Most of the PRE-EARTHQUAKES research activities was devoted to the real-timecomparison and integration of parameters having not onlydifferent space/time resolution and temporal dynamics (i.e., het-erogeneous time-dependent datasets) but also showing differenttime relations to the time of earthquake occurrence.

The 4DEOS potential was exploited both in the learning phase ofthe project - focused on the study of the seismic events occurred inthe past in three selected testing areas (Italy, Turkey and Sakhalin)e and in the experimental phase of the project, named PRIME (Pre-earthquakes Real-time Integration and Monitoring Experiment),carried out over selected areas (Italy, Greece, and Turkey forEurope; Kamchatka, Sakhalin, and Japan for Asia) to perform anactual short-term earthquake prediction by means of a real-timeintegration and analysis of independent observations. About 5000files related to 18 different data/parameters (11 are areal data, 2linear profiles, 5 punctual data) acquired from March 1, 2007 toDecember 31, 2012 were shared, compared, and integrated.

Considering the specific case of the Abruzzo earthquake (April 6,2009, Mw 6.3), the 4DEOS platform was used to compare differentchemicalephysical parameters (measured by ground and satellite

3 ULF is for Ultra Low Frequency.4 VLF is for Very Low Frequency.

systems) in order to identify possible spatial/temporal relationsamong their (even asynchronously occurring) anomalies and theearthquake occurrence. The parameters shared by providers withthe 4DEOS contain information about the measures, in particular ifthey are anomalous or not and a legend containing all details foreach visualized parameter is displayed over the map (i.e., Fig. 4).

Table 1 shows an overview of all measured (ground and satellitebased) parameters for the case of the Abruzzo earthquake. For eachof them, product features (e.g., space-time resolution) and time lag(here reported in terms of number of days) of the observedanomalous value (red table cells in accord with above) with respectto the earthquake occurrence (last column in the table) arereported.

Fig. 4 shows a screenshot of the 4DEOS Client single view modewhere the independent data/products provided by the projectpartners for the Italian testing area are reported. Such data werevisualized in the 4DEOS Client and analyzed in the temporaldomain through the temporal (global and local) bars. To facilitateinterpretation, data/products containing anomalous occurrences ofthe represented parameter (identified on the basis of specific dataanalysis methodologies) are represented by a red symbol (forpunctual and linear features) or contoured by a red line (for arealdata). Looking at Fig. 4, it is possible to note that the thermalanomaly maps obtained by MSG/SEVIRI satellite data (Genzanoet al., 2009) on March 30, 2009 (i.e., 7 days before the earth-quake) could be visualized together with the other anomaloussignals (like, for example, the vertical Total Electron Contentmeasured by GPS17 receivers; see PRE-EQ D.10_2, 2013) whichoccurred up to few hours before the Abruzzo earthquake.

Figs. 5 and 6 show two examples of an earthquake forecast doneusing 4DEOS during the real-time project phase (PRIME). The firstcase (Fig. 5) is related to the Kahramanmaras earthquake (M~5) onJuly 22, 2012 and represents the classical example of superimpo-sition (achievable also by using other GIS systems) of a temporallyvariable layer (TIR anomaly maps) over a static one reportinggeographic references and significant (temporally stable) geotec-tonic settings (mostly faults). In this case, the thermal anomaliesobserved since July 18 up to July 20, 2012 in the Eastern Turkeyincrease their significance because they appear just close to thetectonic lineaments (dashed green line superimposed on TIRanomalies) of the epicenter area a few days before the event.

In the second case (Fig. 6), the importance of the 4DEOS capa-bility to asynchronously visualize multi-parametric geospatial ob-servations is more clearly highlighted. The evident correlationbetween TIR anomalies (appearing in the Aegean Sea just a fewdays before the earthquake of magnitude 5.7 occurred on August29, 2014) and the increase in seismicity (observed exactly in thesame area but starting in a different day) was fully appreciated onlythanks to the 4DEOS visualization system that allows an asyn-chronous comparison between independent geospatial data.

In the framework of the PRE-EARTHQUAKES project the 4DEOSplatform thus proved to be a quite important tool to study earth-quake preparation phases. Thanks to this platform, it was possible:

� to visually analyze asynchronous signals, simultaneously orindividually. Moving in the space-time domain it was possible tobetter understand the relationship between different parame-ters and their link with the earthquake occurrence, both in an a-posteriori analysis (e.g., seismic events occurred in the past) andin a real-time monitoring phase (PRIME);

17 Global Positioning System.

Fig. 4. A screenshot of the 4DEOS Client, where some products used to study the preparatory phases of the Abruzzo seismic event (April 6, 2009, Mw 6.3) are shown. Different products, based both on satellite technologies and groundstations, are superimposed. Note that all the punctual and linear geographical features are linked to the graphical representation of the (space or time) variations of a considered variable, which can be displayed by clicking on them(down left graphic). Note that parameters having a punctual or linear representation are indicated with a red symbol (like plasma frequency in the picture) in case of the presence of anomalies (in blue if in normal conditions, like forGPS-TEC). In the same way, TIR maps containing TIR anomalies are contoured in red to indicate this circumstance (see text). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version ofthis article.)

R.Pacielloet

al./Environm

entalModelling

&Softw

are77

(2016)50

e62

57

Table 1Products shared within the 4DEOS platform for the Abruzzo (April 6, 2009, Mw 6.3) test case during PRE-EARTHQUAKES among the project partners.

Fig. 5. Use of the 4DEOS Client in a classic mode: the temporal local sliding bar related to thermal anomaly (TIR) maps is used to compare TIR map evolution over a static layerreporting faults and tectonic lineaments (solid black and dotted green lines) of the eastern part of Turkey. TIR anomalies detected just close to main faults (between 18 and 20 July,2012) were used to correctly predict (PRE-EQ D.10_2, 2013) the Kahramanmaras earthquake of magnitude 5 which occurred a few days later (on July 22, 2012). Note that anadditional dynamic layer, which refers to earthquake epicenters and magnitudes, is visualized at the time of the Kahramanmaras earthquake (represented by a green star in allscreenshots). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 6. Use of the 4DEOS Client in an actual asynchronous mode: the temporal local slide bars related to TIR maps and to earthquake occurrences are asynchronously adjusted in order to enhance the correspondence between spatialdistribution of foreshocks (22, 23, and 24 August 2014) and TIR anomalies (25 August 2014) preceding the main shock (green circle) which occurred a few days later (August 29, 2014) just in the middle of the area of “convergence” inthe Aegean Sea. It is one example of a successful earthquake prediction allowed by the use of the 4DEOS functionalities, among several cases tested during the PRIME experiment. (For interpretation of the references to colour in thisfigure legend, the reader is referred to the web version of this article.)

R.Pacielloet

al./Environm

entalModelling

&Softw

are77

(2016)50

e62

60

R. Paciello et al. / Environmental Modelling & Software 77 (2016) 50e62 61

� to carry out multi-parametric visual analyses using severaldatasets, which were composed of products having heteroge-neous (spatial and temporal) features and processed usingdifferent methodologies. Results were achieved without theneed to transfer data among the PRE-EARTHQUAKES partners aswell as adopting several levels of authorization (restrictions ongeographic areas) and security policies following the MOMOconcept.

5. Summary and conclusions

The purpose of this paper was to describe an innovative solution(4DEOS e 4 Dimensions Environmental ObServation) for the easymanagement and visualization of heterogeneous and asynchro-nous geospatial datasets.

Thanks to the possibility of setting one or more geospatial time-dependent parameters at a given time, while moving othersthrough time, the 4DEOS platform is able to manage the temporaldimension in an asynchronous way, differently from any other GISsoftware. In particular, the proposed solution allows users tomanage the temporal dimension (t) of geospatial (x,y,z) parametersin order to better understand possible space-time relations amongindependent geospatial datasets.

The 4DEOS platform consists of components that are fully basedon open source software and it is based on a service-orientedClient-Broker-Server architecture, where data shared by providers(throughWMS services, according to OGC standards) are queryableby clients without any need to transfer files, thus preserving dataprovider ownership.

We presented an example of the use of the developed system forthe case of earthquake forecast studies. The 4DEOS potentiality wasverified during the EC-FP7 PRE-EARTHQUAKES project activities,where different physical and chemical observations provided byground and satellite techniques were asynchronously compared toimprove the reliability of a multi-parametric short-term earth-quake forecast system.

The 4DEOS system was used in this context as a common inte-gration and visualization platform and it proved its crucial ability tomanage the temporal dimension in an asynchronous way allowinga successful short term forecast of several earthquakes both in theoff-line (simulating multi-parametric integration consideringevents occurred in the past) and real-time (operational) mode. Theimportance of the adoption of the MOMO principle within 4DEOSwas also demonstrated as the system encouraged wide datasharing and strict collaboration among different partners, includinginstitutions not usually willing to share expensive data or addedvalue products for free.

The 4DEOS is currently being used for the real-time monitoringof Italy, Greece, Turkey, and Sakhalin. Thanks to its capability ofdisplaying and managing heterogeneous and asynchronous data,also at a large scale, the 4DEOS was included among the demon-strators supporting one of the Priority Actions (EQuOS, EarthQuakeObserving System) of GEO (Group on Earth Observations, http://www.earthobservations.org) 2012e2015 Work Plan (GEO, 2014).

Acknowledgments

Author contributions: Paciello R., Coviello I. and Bitonto P.designed and developed the 4DEOS platform; Donvito A. and Tra-mutoli V. dealt with the requirements analysis, the functionalityanalysis and the solutions choice; Filizzola C. managed the systemimplementation for geostationary satellite data; Genzano N. andLisi M. handled the ground data integration; Pergola N. dealt withthe system implementation for polar satellite data and Sileo G.

contributed to write this manuscript.The research leading to these results received funding from the

European Union Seventh Framework Programme (FP7/2007e2013)under grant agreement n� 263502 e PRE-EARTHQUAKES project:Processing Russian and European EARTH observations for earth-QUAKE precursors Studies. The document reflects only the author'sviews and the European Union is not liable for any use that may bemade of the information contained herein.

The research leading to these results was partly funded also byBasilicata Region through the ERDF/NIBS (European RegionalDeveloping Fund/Networking and Internationalization of BasilicataSpace technologies) project and International Space Science Insti-tute (BerneSwitzerland).

References

Abraham, T., Roddick, J.F., 1999. Survey of spatio-temporal databases. Geo-Informatica 3 (1), 61e99.

Adebayo, O., Neilson, J., Petriu, D., 1997. A performance study of client-broker-serversystems. In: Proceeding CASCON '97 Proceedings of the 1997 Conference of theCentre for Advanced Studies on Collaborative Research, pp. 1e15.

Adler, R., 1995. Distributed coordination models for client/server computing. IEEEComput. 14e22.

Bonfanti, P., Genzano, N., Heinicke, J., Italiano, F., Martinelli, G., Pergola, N.,Telesca, L., Tramutoli, V., 2012. Evidence of CO2-gas emission variations in thecentral Apennines (Italy) during the L'Aquila seismic sequence (March-April2009). Boll. Geofis. Teor. Appl. 53 (1), 147e168. http://dx.doi.org/10.4430/bgta0043.

Cicerone, R.D., Ebel, J.E., Britton, J., 2009. A systematic compilation of earthquakeprecursors. Tectonophysics 476, 371e396. http://dx.doi.org/10.1016/j.tecto.2009.06.008.

Fan, Y.T., Yang, J.Y., Zhu, D.H., Wei, K.L., 2010. A time-based integration method ofspatio-temporal data at spatial database level. Math. Comput. Model. 51 (2010),1286e1292.

Frank, A.U., 1998. Different types of “Time” in GIS. In: Egenhofer, M.J., Golledge, R.G.(Eds.), Spatial and Temporal Reasoning in Geographic Information Systems.Oxford University Press, New York/Oxford, pp. 40e62.

Galton, A., 2004. Fields and objects in space, time, and space-time. Spat. Cognit.Comput. 4 (1), 39e68. http://dx.doi.org/10.1207/s15427633scc0401_4.

Gebbert, S., Pebesma, E., 2014. TGRASS: a temporal GIS for field based environ-mental modeling. Environ. Model. Softw. 53, 1e12. http://dx.doi.org/10.1016/j.envsoft.2013.11.001.

Genzano, N., Aliano, C., Corrado, R., Filizzola, C., Lisi, M., Mazzeo, G., Paciello, R.,Pergola, N., Tramutoli, V., 2009. RST analysis of MSG-SEVIRI TIR radiances at thetime of the Abruzzo 6 April 2009 earthquake. Nat. Hazards Earth Syst. Sci. 9,2073e2084.

GEO, 2014. GEO 2012-2015 Work Plan, pp. 1e86. Available. http://www.earthobservations.org/documents/work%20plan/geo_wp1215_rev4_141127.pdf.

Lacava, T., Brocca, L., Calice, G., Melone, F., Moramarco, T., Pergola, N., Tramutoli, V.,2010. Soil moisture variations monitoring by AMSU-based soil wetness indices:a long-term inter-comparison with ground measurements. Remote Sens. En-viron. 114, 2317e2325. http://dx.doi.org/10.1016/j.rse.2010.05.008.

Langran, G., 1993. Issues of implementing a spatiotemporal system. Int. J. Geogr. Inf.Syst. 7 (4), 305e314.

Langran, G., Chrisman, N.R., 1988. A framework for temporal geographic informa-tion. Cartographica 25 (3), 1e14.

Le, Y., 2005. A prototype temporal gis for multiple spatio-temporal representations.Cartogr. Geogr. Inf. Sci. 32 (4), 315e329. http://dx.doi.org/10.1559/152304005775194737.

Le, Y., 2012. Challenges in data integration for spatiotemporal analysis. J. Map Geogr.Libr. Adv. Geospat. Inf. Collect. Arch. 8 (1), 58e67. http://dx.doi.org/10.1080/15420353.2011.622457.

Lee, E., Moon, M., Gunst, R.F., 2006. A spatiotemporal model for pollutant concen-trations. Commun. Stat. Theory Methods 35 (3), 489e497. http://dx.doi.org/10.1080/03610920500476465.

Maceachren, A.M., Wachowicz, M., Edsall, R., Haug, D., Masters, R., 1999. Con-structing knowledge from multivariate spatiotemporal data: integratinggeographical visualization with knowledge discovery in database methods. Int.J. Geogr. Inf. Sci. 13 (4), 311e334. http://dx.doi.org/10.1080/136588199241229.

Marchese, F., Falconieri, A., Pergola, N., Tramutoli, V., 2014. A retrospective analysisof the Shinmoedake (Japan) eruption of 26e27 January 2011 by means ofJapanese geostationary satellite data. J. Volcanol. Geotherm. Res. 269, 1e13.http://dx.doi.org/10.1016/j.jvolgeores.2013.10.011.

Open Geospatial Consortium, 2006. OpenGIS Web Map Server ImplementationSpecification, p. 85. OGC, 06-042.

Open Geospatial Consortium, 2008. Web Coverage Service (WCS) ImplementationSpecification, p. 119. OGC, 07-067r5.

Open Geospatial Consortium, 2005. Web Feature Service Implementation Specifi-cation, p. 112. OGC, 04-094.

Pelekis, N., Theodoulidis, B., Kopanakis, I., Theodoridis, Y., 2004. Literature review of

R. Paciello et al. / Environmental Modelling & Software 77 (2016) 50e6262

spatio-temporal database models. Knowl. Eng. Rev. 19 (3), 235e274. http://dx.doi.org/10.1017/S026988890400013X.

Peuquet, D.J., 2001. Making space for time: issues in space-time data representation.GeoInformatica 5 (1), 11e32.

Peuquet, D.J., 2005. Theme section on advances in spatio-temporal analysis andrepresentation. ISPRS J. Photogramm. Remote Sens. 60, 1e2.

Peuquet, D.J., Duan, N., 1995. An event-based spatiotemporal data model (ESTDM)for temporal analysis of geographical data. Int. J. Geogr. Inf. Syst. 9 (1), 7e24.

PRE-EQ D10_2, 2013. PRE-earthquakes Final Project Report. Available on line in“Downloads” area of. http://www.pre-earthquakes.org/.

Pultar, E., Cova, T.J., Yuan, M., Goodchild, M.F., 2010. EDGIS: a dynamic GIS based onspace time points. Int. J. Geogr. Inf. Sci. 24 (3), 329e346.

Rasinmaki, J., 2003. Modelling spatio-temporal environmental data. Environ.Model. Softw. 18, 877e886.

Sadahiro, Y., 2002. A graphical method for exploring spatiotemporal point distri-butions. Cartogr. Geogr. Inf. Sci. 29 (2), 67e84. http://dx.doi.org/10.1559/152304002782053332.

Scholz, C.H., Sykes, L.R., Aggarwal, Y.P., 1973. Earthquake prediction: a physical basis.Science 181, 803e810. http://dx.doi.org/10.1126/science.181.4102.803.

Shaw, S., Yu, H., 2009. A GIS-based time-geographic approach of studying individualactivities and interactions in a hybrid physical-virtual space. J. Transp. Geogr. 17,141e149. http://dx.doi.org/10.1016/j.jtrangeo.2008.11.012.

Shaw, S., Yu, H., Bombom, L.S., 2008. A space-time GIS approach to exploring largeindividual-based spatiotemporal datasets. Trans. GIS 12 (4), 425e441.

Shi, W., Zhang, M., 2000. Development of a GIS data model with spatial, temporaland attribute components based on object-oriented approach. Geo-spat. Inf. Sci.3 (l), 17e23.

Sugam, S., Shashi, G., Udoyara Sunday, T., 2013. Parametric database approachintegration for handling temporal data in GIS. Geo-spat. Inf. Sci. 16 (2), 91e99.http://dx.doi.org/10.1080/10095020.2013.772807.

Tryfona, N., Jensen, C.S., 1999. Conceptual data modeling for spatiotemporal ap-plications. GeoInformatica 3 (3), 245e268.

Vivar, J.C., Ferreira, M.A.R., 2009. spatiotemporal models for gaussian areal data.J. Comput. Graph. Stat. 18 (3), 658e674. http://dx.doi.org/10.1198/jcgs.2009.07076.

Wang, S., Nakayama, K., Kobayashi, Y., Maekawa, M., 2005. An event-basedspatiotemporal approach. ECTI Trans. Comput. Inf. Theory 1 (1), 15e23.

Zhou, X., Chen, J., Jiang, J., Zhu, J., Li, Z., 2004. Event-based incremental updating ofspatio-temporal database. J. Cent. South Univ. Technol. 11 (No. 2), 192e198.

Zhu, H., Huang, Q., Guo, J., Zhong, S., 2008. An enhanced approach to representingcomplex spatio-temporal data in GIS. In: 2008 International Workshop onEducation Technology and Training & 2008 International Workshop on Geo-science and Remote Sensing, pp. 167e170. http://dx.doi.org/10.1109/ETTandGRS.2008.164.

URL References

http://54north.org/.http://geobatch.geo-solutions.it.http://geoserver.org.http://postgis.refractions.net/.http://udig.refractions.net.http://www.eclipse.org/.http://www.geoapi.org/.http://www.geotools.org/.http://www.opengeospatial.org.http://www.oracle.com/technetwork/java/javase/tech/jai-142803.html.http://www.osgeo.org/.http://www.pre-earthquakes.org/.http://www.vividsolutions.com/jts/JTSHome.htm.https://java.net/projects/imageio-ext.http://www.earthobservations.org.