SR10

Early Warning Systems in Earth Management

Kick-Off-Meeting10 October 2007Technical University Karlsruhe

Programme & Abstracts

GEOTECHNOLOGIENScience Report

No. 10


ISSN: 1619-7399

In addition to currently implemented measures for establishing an early tsunamiwarning system in the Indian Ocean, the German Federal Ministry of Education andResearch (BMBF) has launched a portfolio of 11 research projects for developingand testing early warning systems for other natural geological catastrophes. Theprojects are carried out under the umbrella of the national R&D-Programme GEO-TECHNOLOGIEN.

The overall aim of the integrated projects is the development and deployment ofintegral systems in which terrestrial observation and measurement networks arecoupled with satellite remote sensing techniques and interoperable informationsystems. All projects are carried out in strong collaboration between universities,research institutes and small/medium sized enterprises on a national and interna-tional level.

The abstract volume contains the presentations given at the “Kick-Off-Meeting”held in Karlsruhe, Germany, in October, 2007. The presentations reflect the multi-disciplinary approach of the programme and offer a comprehensive insight into thewide range of research opportunities and applications.

The GEOTECHNOLOGIEN programme is funded by the Federal Ministry for

Education and Research (BMBF) and the German Research Council (DFG)

No.

10Ea

rlyW

arni

ngSy

stem

sin

Earth

Man

agem

ent

GEO

TECH

NO

LOG

IEN

Scie

nce

Repo

rtUmschlag_SR10.qxd 20.09.2007 15:24 Uhr Seite 1


Early Warning Systems inEarth Management



No. 10

Number 1

Impressum

SchriftleitungDr. Ludwig Stroink

© Koordinierungsbüro GEOTECHNOLOGIEN, Potsdam 2007ISSN 1619-7399

The Editors and the Publisher can not be held responsible for the opinions expressed andthe statements made in the articles published, such responsibility resting with the author.

Die Deutsche Bibliothek – CIP Einheitsaufnahme

GEOTECHNOLOGIEN; Early Warning Systems in Earth Management,Kick-Off-Meeting10 October 2007, Technical University Karlsruhe,Programme & Abstracts –Potsdam: Koordinierungsbüro GEOTECHNOLOGIEN, 2007(GEOTECHNOLOGIEN Science Report No. 10)ISSN 1619-7399

Bezug / DistributionKoordinierungsbüro GEOTECHNOLOGIENHeinrich-Mann-Allee 18/1914473 Potsdam, GermanyFon +49 (0)331-620 14 800Fax +49 (0)331-620 14 [email protected]

Bildnachweis Titel / Copyright Cover Picture:NASA/JSC; http://visibleearth.nasa.gov

Preface

Geological events like earthquakes, volcaniceruptions, and landslides devastate manyregions of our planet time and again. Naturalevents increasingly create natural catastrophes– regionally due to dense concentration ofpeople and property in threatened areas, andglobally due to the unification of the worldeconomy. An exact prognosis of where andhow earthquakes occur or volcanoes erupt istherefore essential for effective protection ofthe population and the economy.

In the frame of the R&D-Programme GEO-TECHNOLOGIEN 11 joint projects betweenacademia and industry have been launched in2007. The objective of this research is thedevelopment and deployment of a new gene-ration of early warning systems against earth-quakes, volcanic eruptions and landslides. Alljoint projects are funded by the FederalMinistry of Education and Research (BMBF)with about € 9 Million for the next three years.

Currently supported activities focus on the fol-lowing key themes:1. Development and improvement of measure-

ment and observation systems in real timefor online transmission of decisive physical-chemical danger parameters

2. Development and calibration of coupledprognosis models for quantitative determi-nation of physical-chemical processes with-in and at the surface of the Earth

3. Improvement in the reliability of forecastsand prognoses for decision-making andoptimization of disaster control measures

4. Implementation of mitigation measures inconcrete socioeconomic damage prognoses

5. Development of information systems thatensure prompt and reliable availability of allinformation necessary for technical imple-mentation of early warning and for deci-sion-making by disaster managers

The main objective of the Kick-Off-Meetingwas to bring together the scientists and investi-gators of the funded projects to present theirideas and proposed work plans to each other;several projects are interlinked and could there-fore benefit from synergies. All who are inter-ested in the forthcoming activities of the pro-jects - from Germany, Europe or overseas – arewelcome to share their ideas and results.

Ludwig StroinkMax Wyss

Table of Contens

Scientific Programme Kick-Off-Meeting »Early Warning Systems« . . . . . . . . . . . . . . . . 5

GPS Stress Evaluation within Seconds (G-SEIS)Rothacher M., Gendt G., Galas R., Schöne T., Ge M. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

Rapid Automated Determination of Seismic Source Parameters (RAPID)Meier T., Dahm T., Friederich W., Hanka W., Kind R., Krüger F., Ohrnberger M., Scherbaum F., Stammler K., Yuan X. . . . . . 14

WeraWarn: Real time detection of tsunami generated signatures in currentmaps measured by the HF radar WERA to support coastal re-gions at riskStammer D., Pohlmann T., Gurgel K.-W., Schlick T., Helzel T., Kniephoff M. . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

Early Warning System for Transport Lines (EWS TRANSPORT)Wenzel F., Kühler T., Hohnecker E., Buchmann A., Schedel F., Schöbinger F., Bonn G., Hilbring D., Quante F. . . . . . . . . . . 31

METRIK – Model-Based Development of Technologies for Self-OrganizingDecentralized Information-Systems in Disaster Management (DFG-Graduiertenkolleg)Fischer J., Avanes A., Brüning S., Fahland D., Gläßer T. M., Köhne K., Quilitz B., Sadilek D. A.

Scheidgen M., Wachsmuth G., Weißleder S., Kühnlenz F., Poser K. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

Earthquake Disaster Information System for the Marmara Region, Turkey (EDIM)Wenzel F., Erdik M., Zschau J., Milkereit C., Redlich J. P., Lupp M., Lessing R., Schubert C. . . . . . . . . . . . . . . . . . . . . 51

Numerical Last-Mile Tsunami Early Warning and Evacuation Information SystemBirkmann, Dech, Hirzinger, Klein, Klüpfel, Lehmann, Mott, Nagel, Schlurmann, Setiadi, Siegert, Strunz . . . . . . . . . . . . . 62

Sensor based Landslide Early Warning System – SLEWSArnhardt C., Asch K., Azzam R., Bill, R., Fernandez-Steeger T. M., Homfeld S. D., Kallash A., Niemeyer F.,

Ritter H., Toloczyki M., Walter K. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

Integrative Landslides Early Warning Systems (ILEWS)Glade T., Becker R., Bell R., Burghaus S., Danscheid M., Dix A., Greiving S., Greve K., Jäger S., Kuhlmann H.,

Krummel H., Paulsen H., Pohl J., Röhrs M. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

Development and testing of an integrative 3D early warning system for alpineinstable slopes (alpEWAS)Thuro K. , Wunderlich T., Heunecke O. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

Development of suitable information systems for early warning systemsBreunig M., Reinhardt W., Ortlieb E., Mäs S., Boley C., Trauner F. X., Wiesel J., Richter D.,

Abecker A., Gallus D., Kazakos W . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

Exupéry: Managing Volcanic Unrest – The Volcano Fast Response SystemHort M., Wassermann J., Dahm T. and the Exupéry working group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124

Authors’ Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134

GEOTECHNOLOGIEN Science Report’s – Already published/Editions . . . . . . . . . . . . . . . 136

5

Scientific Programme Kick-Off-Meeting»Early Warning Systems for Natural Hazards«10 October 2007

8.30 Welcome

8.40–9.00 GPS Stress Evaluation within Seconds (G-SEIS)

9.00–9.30 Rapid Automated Determination of Seismic Source Parameters (RAPID)

9.30–9.50 Real time detection of tsunami generated signatures in current maps(WeraWarn)

9.50–10.10 Early Warning System for Transport Lines (TRANSPORT)

10.10–10.30 Presentation DFG-Graduiertenkolleg »METRIK« ModellbasierteEntwicklung von Technologien für selbstorganisierende dezentraleInformationssysteme im Katastrophenmanagement

10.30–11.00 Coffee break

11.00–11.40 Earthquake Disaster Information System for the Marmara Region, Turkey(EDIM)

11.40–12.20 Numerical last-mile Tsunami Early Warning and Evacuation InformationSystem (Last Mile/T-RISK)

12.20–13.20 Lunch

13.20–13.50 Development of a geoservice infrastructure as basis for early warningsystems for landslides (SLEWS)

13.50–14.20 Integrative Early warning System for landslides (ILEWS)

14.20–14.40 Development and testing of an integrative 3D early warning system foralpine instable slopes (alpEWAS)

14.40–15.10 Coffee break

15.10–15.40 Development of suitable Informationssystems for EWS (EGIFF)

15.40–16.20 Managing Volcanic Unrest-The Volcano Fast Response System (Exupery)

16.20–17.00 Final discussion

7

G-SEIS: GPS Surface Deformationswithin Seconds

AbstractThe G-SEIS (Gps-SurfacE deformations withinSeconds) focuses on the monitoring of theEarth surface on different temporal and spatialscales using GPS and/or similar or futurespace-based GNSS systems (GLONASS, GALI-LEO) for the new generations of monitoringsystems for hazards like earthquakes, landslides and volcano eruptions. In order to meetthe requirement on accuracy, time latency,spatial and temporal resolution of the GPSderived deformation, new conceptual sensorstation with modern data communication andtransfer capability will be developed. Newstrategies of real-time data processing for dif-ferent applications are designed and the relat-ed software package will be developed forreal-time data processing of huge global net-works with both static and kinematic stations.The developed system will be tested in twoprototype applications, GPS-Shield for Sumatraand Automatic GPS Array for Volcano EarlyWarning System (AGAVE). The related workpackages are presented here in details.

1. IntroductionWhile GPS is able to measure and quantify sur-face movements of a source region in real-time, seismological networks need a certaintime span to accumulate data in order todetermine the nature of the rupture processand some remote sensing techniques with ahigh spatial sampling give only inadequatetemporal sampling depending on orbital pa-rameters of the satellites. For time-criticalapplications, as e.g. in the case of the tsuna-

mi event in Sumatra, GPS networks might bethe only way to get a fast insight into thenature of a rupture process. Using combinedseismological and GPS information will enableto issue alerts and also to support the nearreal-time modeling of tsunami propagation.Today, in scientific applications for Earth mon-itoring GPS is mainly used in continental-scaleand regional, but data is processed in batches,leading to certain latency in providing positionresult. Therefore, the current status of the GPSmonitoring stations and analysis software cannot fulfill the future requirement on GPS forthe use in, e.g. a tsunami early warning sys-tem or volcano observatories.The G-SEIS (Gps-SurfacE deformation withInSeconds) project focuses on the monitoring ofthe Earth surface on different temporal andspatial scales using GPS and/or similar orfuture space-based GNSS systems (GLONASS,GALILEO) for a new generations of monitoringsystems for hazards, like earthquakes, landslides and volcano eruptions. This confronts uswith the following tasks:– Design and development of a new series of

multi-parameter stations running unattend-ed and autonomously for long periods,

– Development of new and automatedGPS/GNSS software allowing real-time dataanalysis and self-detection of events, and

– Development and test of new data commu-nication strategies for high-rate and high-volume data.

Within the project, a new generation ofrobust, autonomous field GPS-based monitor-ing stations, having modular software and

Rothacher M., Gendt G., Galas R., Schöne T., Ge M.

Department of Geodesy and Remote Sensing

GeoForschungsZentrum Postsdam

14473, Potsdam, Germany

8

hardware architecture, will be developed.Hybrid power subsystems for these field sta-tions will be designed and developed. Varioussatellite- and wireless terrestrial data commu-nication technologies for high- and very high-rate GPS data will be suggested. Appropriatesoftware tools for high-rate (1Hz) real-timedata streaming will be developed. The per-formance of the station’s equipment in diffi-cult environment will then be verified.A real-time software packages will be devel-oped, which will be capable of:– Providing satellite orbits and clocks in real-

time (network software),– Kinematic monitoring of station positions in

real-time for networks with good real-timedata transmission, and

– Performing real-time Precise Point Po-sitioning (PPP) for stations with problems inhigh-rate data communications based onorbit and clocks from the network software(in such cases only orbits and clocks have tobe transferred to the station).

Special studies will yield solid background forfuture application of

– High-rate GNSS data, where high-ratemeans in this context 20 to 50 Hz in thiscontext,

– GNSS combined solutions comprising GPS,GLONASS and the upcoming Galileo system,which will yield a new quality for manyapplications by offering the possibility ofusing nearly 100 satellites simultaneously.

Two prototype deformation monitoring net-works will be set-up and operated. The proto-types should demonstrate that the main goalof the undertaking, i.e. the detection of theEarth’s deformation and issuing warningalarms is possible »within seconds«.

2. Sensor Stations, Data Transferand CommunicationConverting scientific monitoring networks intoearly warning networks requires newapproaches in terms of autonomous opera-tion, failure tolerance, error recovery, powersupply and data communication. During thepast 10 years GFZ has developed GPS field sta-tions for different projects (global tectonicmonitoring, volcano monitoring). Based on

Figure 1: Monitoring station utilizing different techniques (Sensors, Communication, Power Supply)

9

this experience new station concepts will bedeveloped, providing new types of multi-sen-sor (smart) stations.Working with a laptop connected via a100MBit/s internet connection makes it hardto imagine transfer rates of 2400 Bit/s. How-ever, many remote areas have a very limiteddial-up access to global telecommunication.This requires a cascade of measures to e.g.down-sample or compress data streams, todevelop concepts of emergency dial-ups andto test new approaches of data transfer meth-ods. VSAT, BGAN, HF or L-Band communica-tion are among them to be studied. Part of theproject is to test the different devices and todevelop adapted data transfer techniques.A new requirement is also the multi-sensor/multi-purpose approach. Since the installationand maintenance of remote stations are veryexpensive, stations needs to be equippedwith different sensors. For the G-SEIS projectthe focus will be on the combination ofGPS, seismic, tide gauges and weather sen-sors. To achieve a reliable station operation,appropriate hardware has to be selected andnew station management software has to bedeveloped.Also of immense importance is the power sup-ply. Only very few stations have mains power.For remote areas monitoring stations relymostly on solar power. Sophisticate conceptswill be developed reliable power supply andmonitoring tools.

3. Software DevelopmentThe software packages (e.g. BERNESE, GAMIT,EPOS) available for global and regional GPSdata analysis are based on batch processing,i.e. the analysis of data in daily, hourly or evenshorter batches. This strategy has its naturallimitation, if applications are approaching real-time monitoring. Although it is feasible togenerate precise orbit predictions (except incase of maneuvered satellites), clock predic-tions with sub-nanosecond accuracy are notpossible.Commercially available real-time softwarepackages are normally designed under theassumption that phase biases can be precisely

represented by those of nearby reference sta-tions. They are suitable for local/regional net-works, but cannot handle global analysis prob-lems. An alternative, but most promising strat-egy for the global monitoring applications isthe planned real-time IGS service. The globaldata shall be analyzed in real-time for provid-ing real-time orbits and clocks. Based on thosereal-time products with highest possible accu-racy, the position of a single station isobtained by PPP. In order to obtain highly pre-cise and reliable position results, online quali-ty control is one key element of the strategy.Proper integer ambiguity-fixing is a challengeand will improve the PPP accuracy dramatical-ly. First tests will be performed.

3.1. Real-Time Softwarea) Generation of orbits and clocks in real-time:All data analysis is based on data streams froma global GNSS network with possible regionaldensification. A tool has to be set up for cen-tralized data acquisition and pre-processing,which handles all possible events and providespre-processed data for the selected network inthe given sampling rate (e.g., 1 Hz) to theanalysis modules.A special analysis line will be set up using datafrom a global network with best data com-munications (GFZ project sites and sites fromthe IGS for global applications and EUREF forEuropean applications) to provide real-time(latency should be less than 5–10 sec) satelliteorbits and clocks. In a first stage the softwarewill deliver satellite clocks for given predictedsatellite orbits only.Tests will be performed whether the uncali-brated phase delay (UPD), originating in satel-lites, can be estimated reliably from the glob-al network to enable ambiguity-fixing withinPPP for given user stations.Based on the generated orbits and clockslocal and regional station networks will bemonitored.b) Station network solution in real-timeFor local or regional networks with optimaldata connections, applied for earthquakedetection and volcano monitoring, a continu-ous monitoring will be set up. Studies will be

10

performed whether a classical network solu-tion or a PPP plus network-ambiguity-fixing isthe best suitable approach for an efficient andvery precise relative positioning applicable forthe prototype applications in Section 5.

3.2. Real-Time Software: PPP ModeFor special applications, where high-rate GPSdata cannot be transmitted to an analysis cen-ter, e.g. GPS on buoys, in case of an event, theorbit and clock products from global solutionswill be transmitted to the affected stations incase of an event, and the analysis will imme-diately start on the local computer in a PPPmode. To get an optimal time series solutionthe analysis will start with data from severalhours backward (<24 hours) and proceedafterwards with the real-time monitoring untilthe end of the event, and thus will be able todetect tsunami waves (in case of buoys) oranalyze post-seismic behavior in case of earth-quakes or volcano activities.To support such applications a real-time PPPsoftware tool will be developed. Tests will beperformed whether integer ambiguity-fixingcan be realized in real-time PPP based on addi-tional information obtained from the network.

3.3 Deformation MonitoringFinally, the obtained solutions will be used tomonitor the station motions in a selectedregion. Automated procedures will make thedecision whether an event has been detectedand consequently trigger alerting sequences.The deformation information of the GPS sta-tions will be tested for its suitability for extrac-tion of earthquake parameters and informa-tion about the earthquake mechanism andrupture process.It should be mentioned that the software willbe able to analyze all GNSS data (GPS,GALILEO and GLONASS), aiming for highestaccuracy.

4. Supporting Studies

4.1. High-Rate Data RatesFor monitoring of tectonic or volcanic motionsand detecting events 1 Hz data are normally

sufficient. However, to get more informationon the source mechanisms a much highersampling is of great interest. It is planned toperform studies with sampling rates of 10 Hzeven 20 Hz. The reliability of the data stream,the quality of the data and the accuracy of theanalysis results will be studied.

4.2. GNSS StudiesThe number of satellites observed simultane-ously is significantly influencing the quality ofthe results. Presently, normally 8 to 12 GPSsatellites are seen simultaneously (dependingon elevation cutoff and latitude). With thecompletion of the GLONASS system and theupcoming GALILEO this number will increaseto about 24 to 35, which will give a new qual-ity in results. Studies about the influence onthe accuracy of real-time results are planned.

5. Prototype ApplicationsOne of our primary tasks will be the imple-mentation of the new hardware and softwareenvironment into applications and warningregimes. Recent discussions and the experi-ence of the Thailand GPS monitoring of theSumatra earthquake showed the advantage ofnot only monitoring the tectonic situation butalso to monitor the station movement in real-time. A fast output of coordinate changeswould enable the real-time modeling of earth-quake source mechanisms and allow real-timepredictions of tsunamigenic hazard potential.Therefore, the general concept will be demon-strated for the region of Sumatra. Hardware isinstalled in the course of GITEWS, thus theseexpertise and resources will be used. Theresults will be made available to the to-be-established Tsunami Warning Center and canbe used as additional information in the warn-ing chain.A second application will be the monitoring ofa volcano. In Indonesia, the Merapi is highly atrisk and already monitored with different sen-sors by GFZ and other institutes. GPS stationshave been operated by the MERAPI project for3 years, but the sites were not permanentlyoccupied. In Mexico the applicants had alsoinstalled and operated a volcano monitoring

11

network. This effort, however, was stoppedafter 1.5 years due to man-power restrictions.Either the monitoring array in Mexico will beresumed, updated and equipped with real-time infrastructure or a new real-time GPS arraywill be set up on an active volcano in Chile.With both applications we will be able todemonstrate directly with in-situ data thepotential of this new technique.

5.1. Prototype Application: GPS Shieldfor SumatraThe main goal of this work package is theinvestigation whether the proposed new strat-egy of a GPS shield, based on real-time con-tinuous GPS, can be successfully applied forTsunami Early Warning Systems as it wasrecently suggested (Sobolev et al., 2006). Inthis concept the deformation information,especially the deformation gradient within aGPS array, is used to obtain more reliableearthquake parameters. Figure 2 shows theGPS derived high-rate position change of sta-tion JOG2 caused by an earthquake.

The Mentawai islands and the West Sumatracoast will be selected for a prototype installa-tion and operation of GPS arrays, because:– It is expected that one of the future tsuna-

migenic earthquakes may occur close to thelarge city Padang.

– A reliable prediction of tsunami wave heightsfor Padang cannot be provided using traditio-nal earthquake-magnitude-based methodsbecause wave heights on the coast of Padangmay differ by more than a factor of 5 forearthquakes having the same seismic magni-tude but different slip distribution. It is expec-ted that the slip distribution can be success-fully measured with small scale GPS arrays.

– Several other large cities, which are alsolocated close to seismogenic zones, also can-not be protected using only seismologicalnetworks. However, GFZ is engaged in thedevelopment of GITEWS and the proposedarray near Padang is of great importance tosuch efforts.

Key elements will be arrays of GPS-basedmulti-sensor stations. High-rate (1–10 Hz) con-

Figure 2: GPS derived high-rate deformation by earthquake.

12

tinuous GPS arrays should be installed onMentawai islands (Siberut and Pagai), alongthe trench, and one or two, on the west coastof Sumatra (near Padang and to the South).The arrays will provide displacement gradientsand precise coordinates of the Array MasterStations. Time series of those parameters willbe used to predict heights of tsunami wavesfollowing the earthquake.

5.2. Prototype Application: Automatic GPSArray for Volcano Early Warning System(AGAVE)The purpose of this work package is the instal-lation and continuous operation of an auto-matic mobile array of GPS receivers, and aux-iliary sensors, to monitor deformations of anactive volcano prior to or during eruptionsand, thus, to provide short-term hazard assess-ments. Continuous GPS provides an excellenttool to set up an early warning system forforecasting and/or prediction of volcanic erup-tions, but necessarily observational data mustbe streamed and processed in real-time. Astate-of-the-art small scale deformation moni-toring system using GPS has been designedand developed at GFZ in the past years. It isan open system, which can be easy extended,but it must be upgraded and equipped withreal-time capability. Implementation of real-time processing, array upgrade and demon-stration of its usefulness for volcano monitor-ing are the main points of this work.Installations of technical equipment on activevolcanoes are critical in terms of working safe-ty. Therefore, alternative project sites are pro-posed. The array will be installed either on anactive volcano in Chile or our previous systemin Mexico (Popocatepetl/Colima) will beupgraded and taken into operation again. Itwill consist of one »Array Master Station«(AMS) and of three or four monitoring »ArraySlave Stations« (ASS). Local partners will sup-port installation and operation of the system.

6. Schedule and WorkflowThe sensor station development will have avery high priority during the first year. Basedon previous experience the hardware will be

tested for use in hazard monitoring projects,and new components will be implemented.This work package is an important extensionof our station management concept.The data transfer and communication will dealwith testing different media for data transfer. Itis important for the future concept of reliabledata transfer. VSAT, internet and classical satel-lite telephone lines will be tested in the firstyear. BGAN (Broadband Global Area Network)will be tested as soon as it is available in the tar-geted region of our prototype applications.The software development will begin withstart of the project. In the first 18 months thecore modules for the real-time software will bewritten and tested (real-time data acquisition,preprocessing, update of GNSS clocks). In thesecond half, the software will be extend forthe processing of a full set of parameters,including orbits and station positions, andusing different data types (fast moving sta-tions, network stations). Deformation monitor-ing for groups of points and the alerting capa-bility will be implemented.The supporting studies will run in parallel tosoftware development during the last year. Assoon as results are available, they will beimplemented and tested. It will certainly needthe new stations developed in the sensor sta-tion. For the first studies no real-time softwareis needed.The main hardware elements of the array willbe set up at GFZ first. All the software mod-ules will be implemented, tested and updated.The engineering model will be used for proto-type The first array will be installed in an easyaccessible area with high risk for hazards.Accessibility is an important issue in the testphase. Most probably it will be located on theisland Pagai (GPS shield Sumatra, Indonesia).The demonstration phase is foreseen for threemonths. The other elements of the prototypeapplications will be installed and taken intooperation afterwards.

7. SummaryGFZ has already demonstrated its capability ofhigh-quality GPS processing and operationalsevice of real-time GPS stations for various

13

applications. The G-SEIS project will modernizethe two already existing chains, namely real-time data transfer and high-precise GPS dataprocessing which are running now separately,and join them together in order to providereal-time, precise and high-rate surface defor-mation for the new generation of hazardmonitoring system. The details of the fivemajor work packages are described as well asthe related new approaches. The approachesdeveloped in the project will, in a very shorttime, revolutionize monitoring of Earth surfacedeformation parameters and detection of haz-ardous earthquakes, volcanoes and landslide.

References:Blewitt, G., et al. (2006): Rapid Determinationof Earthquake Magnitude using GPS forTsunami Warming Systems. Proceedings IGSWorkshop 2006, ESA, Darmstadt, in print.

Bock, H., Beutler, G., Schaer, S., Springer, T.A.,Rothacher, M.: Processing Aspects Related toPermanent GPS Arrays, Earth Planets Space,52, 657–662, 2000.

Galas, R., Ch. Reigber, G. Michel, 1998. Cur-rent Status & Perspectives of the GFZ GPS-Based Volcano Monitoring System, WPGM1998 (invited paper), EOS supplement.

Galas, R., J. Wickert and W. Burghardt: HighRate Low Latency GPS Ground Tracking Net-work for CHAMP, Phys. Chem. Earth (A), 26(6–8), 649–652, 2001.

Genrich, J.F., Y. Bock (2006), Instantaneousgeodetic positioning with 10–50 Hz GPSmeasurements: Noise characteristics andimplications for monitoring networks, J.Geophys. Res., 111, B03403, doi:10.1029/2005JB003617.

Miyazaki, S., K.M Larson, et al.: Modelling therupture process of the 2003 September 25Tokachi-Oki (Hokkaido) earthquake using 1-HzGPS data. Geophysical Research Letters. 2004NOV 16; 31(21): L21603 1–4

Reigber, Ch., R. Galas, W. Koehler, M. Forberg,M. Ramatschi, and J. Wickert:, 2003. GFZHR/LL GPS Ground Station Networks and theirUse, AGU Fall, San Francisco, December 8–12,2003.

Schaer, S., Beutler, G., Rothacher, M., Brock-Mann, E., Wiget, A., Wild, U.: The Impact ofthe Atmosphere and Other Systematic Errorson Permanent GPS Networks, IAG-Sym-posia, Springer Verlag, 121, 373–380, 2000.

Sobolev, S.V., A.Y. Babyeko, R. Wang, R.Galas, M. Rothacher, D.V. Sein, J. Schröter,J. Lauterjung, C. Subarya,: GPS-Shield: reliableprediction of tsunami amplitude within lessthan 10 minutes of an earthquake (acceptedto EOS) 2006.

14

Rapid Automated Determinationof Seismic Source Parameters (RAPID)

IntroductionAutomated rapid and robust detection andlocalization of earthquakes using real-time dataof regional and global networks lies at theheart of any earthquake early warning system.Early recognition of earthquakes and the deter-mination of their source parameters are essen-tial for a rapid warning against earthquakegenerated tsunamis as well as for any rapidpost-earthquake response decisions by emer-gency management authorities. Because ofrecent progress in real-time data transfer aswell as in off-line data processing, automatednear-real time estimation of source parametersis feasible. However, methods have to beadopted for near real-time data processing andsoftware has to be developed that makes useof the incoming real-time data stream. Buildingon existing real-time data transfer and dataprocessing software packages like SeedLinkand SeisComP developed at GFZ Potsdam,

additional software components for automateddetermination of source parameters and a nearreal time estimate of the spatial distribution ofstrong shaking from regional and teleseismicinformation will be developed.The SeedLink IP data transmission protocol aspart of the SeisComP data acquisition and pro-cessing software package has become a de-facto global standard adopted as a manufac-turer independent real-time data exchangeprotocol by international organisations likeFDSN, ORFEUS and IOC/IOTWS. Due to its sim-plicity and robustness, SeisComP is meanwhileused at many institutions worldwide for real-time data acquisition and data processing. TheGEOFON data centre at GFZ acquires present-ly near real-time data from 45 of its 54 sta-tions together with the data from more than250 globally distributed stations from GEOFONpartner networks. Using these real-time datafeeds, automatic processes for data quality

Meier T. (1), Dahm T. (2), Friederich W. (3), Hanka W. (4), Kind R. (5), Krüger F. (6), Ohrnberger M. (7),

Scherbaum F. (8), Stammler K. (9), Yuan X. (10)

(1) PD. Dr. Thomas Meier, Institut für Geologie, Mineralogie und Geophysik, Ruhr-Universität Bochum,

44801 Bochum, Universitätsstr. 150, NA 3/173, [email protected]

(2) Prof. Dr. Torsten Dahm, Universität Hamburg, Institut für Geophysik, Bundesstr. 55, 20146 Hamburg,

[email protected]

(3) Prof. Dr. Wolfgang Friederich, Ruhr-Universität Bochum, Institut für Geologie, Mineralogie und Geophysik,

NA 3/165, Universitätsstr. 150, 44801 Bochum, [email protected]

(4) Dr. Winfried Hanka, GFZ Potsdam, Telegrafenberg, 14473 Potsdam, [email protected]

(5) Prof. Dr. Rainer Kind, GFZ Potsdam, Telegrafenberg, 14473 Potsdam, [email protected]

(6) Dr. Frank Krüger, Universität Potsdam, Institut für Geowissenschaften, PSF 601553, 14415 Potsdam,

[email protected]

(7) Dr. Matthias Ohrnberger, Universität Potsdam, Institut für Geowissenschaften, Karl- Liebknecht Str. 24/25,

14415 Potsdam, [email protected]

(8) Prof. Dr. Frank Scherbaum, Universität Potsdam, Institut für Geowissenschaften, Karl- Liebknecht Str. 24/25,

14415 Potsdam, [email protected]

(9) Dr. Klaus Stammler, Seismologisches Zentralobservatorium Gräfenberg, Bundesanstalt für Geowissenschaften und

Rohstoffe, Mozartstr. 57, 91052 Erlangen, [email protected]

(10) Dr. Xiaohui Yuan, GFZ Potsdam, Telegrafenberg, 14473 Potsdam, [email protected]

15

checks, event detection, localisation andsource quantification are applied and theresulting earthquake information is automati-cally published in the Internet (http://www.gfz-potsdam.de/geofon/new/eq_inf.html) andalerts are distributed by email and SMS mes-sages to a wide spread user community. TheSeisComp seismological control centre soft-ware architecture will also serve as basis forthe application of the software modules devel-oped by the RAPID project.Based on methodical studies, algorithms forautomated near-real time determination ofsource parameters will be developed and test-ed using data available at the GEOFON datacentre. The algorithms will be implemented atthe GEOFON data centre. The developed soft-ware components will complement the exist-ing ones. The software will be portable toother data centres. Tests of the software atthese data centres as well as integration ofgenerated alerts into European early warningsystems are intended.The project consists of 5 work packages. Eachwork package aims at providing a softwarecomponent for a specific purpose. We willfocus on (1) signal detection, signal character-isation and event localisation, (2) rupture mon-

itoring, (3) determination of source mecha-nisms, (4) teleseismic estimation of shake-mapestimations and (5) on evaluation and classifi-cation of source parameter estimates. A flow-chart of the processing scheme we will workon is given in Fig. 1.Broad-band waveforms transferred to theGEOFON data centre in real time are input tothe software package to be developed in theframework of the project. Source parametersare estimated in near-real time and are fedinto several processing schemes. In one pro-cessing chain, the tsunami hazard potential isevaluated based on source parameter esti-mates. In a second chain source parameterestimates are used in conjunction with strongground-motion models to derive first estimatesof the expected spatial distribution of strongshaking (»tele-shakemaps«). In the case ofdetecting a large earthquake or the possibleoccurrence of a tsunami, alert messages aregenerated and further non-seismological haz-ard evaluations are triggered. The algorithmsshould be automated, rapid and robust. Therobustness of the tools has to be intensivelytested, classification schemes of the estimatedsource parameters, and rapid hazard estima-tion algorithms have to be developed in order

Figure 1: Flowchart of the early warning data processing scheme.

16

to generate informative alerts that may beused for further decisions in case of a strongearthquake. The software components will betested using data of events in Indonesia andthe Mediterranean.In the following short descriptions of the 5work packages are given. The project startedon July 1, 2007 and will end on September30, 2010.

Work package 1: Signal detection, signalcharacterisation and event localisation

T. Meier (Ruhr-University Bochum), W. Friede-rich (Ruhr-University Bochum), F. Scherbaum(University Potsdam), K. Stammler (Seismologi-sches Zentralobservatorium Gräfenberg, BGR)

Work package 1 is focused on rapid automat-ed event detection, localisation, and estima-tion of magnitude. Rapid automated eventdetection and localisation are the basis for anyearthquake early warning system. Carried outcontinuously and in near-real time, they repre-sent the first processing steps of any near-realtime data processing chain and triggers subse-quent processing steps if the potential for alarge earthquake is detected. Existing softwarecomponents of the localisation scheme devel-oped at the GEOFON data centre are to besupplemented by additional tools. Crucial forthe accuracy of the localisation is the qualityof automatic arrival-time estimates. Methodsfor the determination of arrival-times based onhigher-order statistics and autoregressive (AR)models will be adapted to near-real time dataprocessing. Special emphasis will be put onthe estimation of S-wave arrival times at localand regional distances that allow the reduc-tion of the uncertainty in event localisations.The potential of a modified algorithm for thedetermination of AR-parameters applied tothree-component waveforms will be investi-gated. The programme HYPOSAT (Schweitzer,2001) will be adapted for rapid event localisa-tion using only a limited number of stations atlocal and regional distances. Additional infor-mation like quality estimates of automaticpicks and polarisation estimates are effectively

accounted for by HYPOSAT. Besides rapiddetection and localization, a fundamentalchallenge to any early warning system is therapid recognition and characterisation ofstrong earthquakes. After a preliminary locali-sation has been obtained, incoming wave-forms will be continuously compared to wave-forms of master events by a cross correlationtechnique. A database of master-event wave-forms will be set up using synthetics (Frie-derich and Dalkolmo, 1995) as well as wave-forms of previous events recorded at GEOFONstations. Similarities to master events will bequantified allowing the rapid identification ofearthquakes with high hazard potential. Incases of a high correlation with master events,source time functions are determined using aWiener filter approach. Magnitude estimatesare obtained from determined source timefunctions.Inputs are the incoming real-time data fromseismological broad-band stations. Outputsare event detections, localizations, magnitudeestimates, and similarities to master events.

Work package 2: Near real time rupturemonitoring

R. Kind (GFZ Potsdam), F. Krüger (University ofPotsdam), X. Yuan (GFZ Potsdam), W. Hanka(GFZ Potsdam), T. Meier (Ruhr-UniversityBochum)

Recently it was shown that it is possible tomonitor the rupture propagation of strongearthquakes with the aid of teleseismic arraytechnology (Krüger and Ohrnberger, 2005; Ishiiet al., 2005). Presently we are working on atechnique to monitor in real time the rupturepropagation using a polarization analysis oflocal and regional events. This is a second inde-pendent technique providing redundancy andit is also closer to real time than teleseismictechniques. Determination of backazimuth andincidence angles from the amplitude ratios ofthe three components of the P phase at onestation is a very old technique for estimation ofthe earthquake epicenter. We apply in real timea moving window technique for the determi-

17

nation of the backazimuth as a function oftime. With the epicenter known from othertechniques, we will monitor the rupture prop-agation starting at the epicenter. We apply thistechnique to a number of stations and get areliable image of the rupture propagation innear real time. Using many stations increasesthe reliability of the observations. We haveapplied this technique to the data of the greatSumatra event from December 2004 andobtained results very close to those of the tele-seismic array technique. The next step will bethe application of the new technique to aseries of other great earthquakes.We intend to develop effective software forautomatic real time application of this tech-nique to any large earthquake occurring any-where on Earth. We also intend to publish theresults in the internet, in a similar way Harvardis publishing the moment tensor solutions, butwe plan to do this in near real time. Theadvantage of our technique is the inclusion ofthe nearest stations in the solution, which per-mits solutions which are faster than only tele-seismic solutions.Inputs are real time seismic broadband datafrom existing seismic stations. Outputs are realtime determination of lengths of ruptures oflarge earthquakes, which is very important fordetermination of the earthquake magnitudeand estimation of possible damages, includinggeneration of tsunamis.

Work package 3: Estimation of sourcemechanisms – kinematic and dynamicsource parameters

T. Dahm (University of Hamburg), F. Krüger(University of Potsdam), T. Meier, W. Friederich(Ruhr-University Bochum), K. Stammler(Seismologisches ZentralobservatoriumGräfenberg, BGR)

Rapid, automatic and robust estimation ofparameters of the extended seismic source likeits duration, its spatial extent, its slip and seis-mic energy release are a necessary prerequisiteto estimate with modern methods the localdamage and more widespread consequences

of strong earthquakes (like tsunami genera-tion) for a specific region of the Earth. Weplan to determine these important quantitieson a global scale from regional and teleseismicbroadband waveforms within 10 to 20 min-utes of the onset of a specific event for eventswith Mw > 5.7. For that purpose we will usehigher order moment tensor inversion on thebase of simple parameterizations of the finitesource (Dahm and Krüger, 1999; Vallee andBouchon, 2004) with constraints provided bywork packages 1 and 2 and the recently devel-oped imaging of the rupture process by arraysin the teleseismic distance range (Krüger andOhrnberger, 2005a, 2005b).Inputs are global/regional seismological broad-band waveforms provided by GEOFON andother data centers, station parameter databas-es and Greens functions from own Greenfunction databases. If available, continuousGPS data will be added. Additional inputs arethe location (first onset, including an estimateof the source depth, initial estimate of scalarmoment and type of mechanism from workpackage 1) and a first estimate of the rupturedirection from work package 2. Outputs areestimates of the source mechanism, scalarmoment and centroid depth, the lateraldimension of the source and its temporalduration, the average rupture velocity, theaverage type of rupture (unilateral/bilateral),the rupture direction and estimates of averageslip and energy.

Work package 4: Near real timeestimation of expected spatialdistribution of strong ground-motion

F. Scherbaum (University Potsdam), F. Krüger(University Potsdam)

ShakeMaps are representations of groundshaking produced by an earthquake which arebecoming an increasingly important tool tofacilitate the rapid post-earthquake responseby private and public agencies. In their origi-nal form they are based on regionally derivedestimations of strong ground-motion fromaccelerometric networks, which are subse-

18

quently interpolated spatially and convertedinto instrumental intensity maps. This approach,however, is limited to well instrumented re-gions. We intend to develop and implement anovel approach to rapidly obtain first estimatesof the expected spatial distribution of strongshaking for strong earthquakes in continentalregions (and with lesser accuracy also for sub-duction zone events) based on teleseismicrecordings (»TeleShakeMaps«). TeleShakeMapswill complement traditional ShakeMaps (Waldet al., 1999) while for remote areas (e.g., thePakistan earthquake of fall 2005) they will pro-vide unique information. The backbone of Tele-ShakeMaps will be the combination of teleseis-mically determined source dimensions (Dahmand Krüger, 1999) obtained in work package 3with composite ground motion models (Scher-baum et al., 2005), calibrated for the prospec-tive application regions using the method ofScherbaum et al. (2004). This will allow notonly to rapidly predict site specific strongground-motion parameters but also to quanti-fy the associated epistemic uncertainties.Inputs are the estimate of source size, sourceorientation, and moment magnitude obtainedfrom work package 5.3 using the approach ofDahm and Krüger (1999) and target regionspecific ground motion models. The latter aregoing to be either empirical (cf. Douglas,2004), stochastic (Boore, 2003), or hybridempirical (Campbell, 2003) using the methodof Scherbaum et. al. (2006) to determine theregional transfer functions. Output is an esti-mate of the spatial distribution of strongground shaking together with associateduncertainty estimates.

Work package 5: Evaluationand classification of near-real time sourceparameter estimates for tsunami hazard

Matthias Ohrnberger (University of Potsdam),Frank Scherbaum (University of Potsdam)

The objective of this work package is thedevelopment of an automatic and robust alertlevel for tsunami generation based on the evi-dence of seismological observations as

described in work packages 1 to 3. The alertlevel system is intended to serve as a prototypenon-public message generator for scientificstaff involved in tsunami warning centers. Thealert level system will be based on the formu-lation of belief networks (Aspinall et al., 2003;Woo and Aspinall, 2005). Belief networksallow incorporating both discrete and contin-uously distributed data sources as well as thequantification of expert knowledge into aprobabilistic framework which can be evaluat-ed very efficiently.For the establishment of such a system it isnecessary to account for all available priorinformation (Moreno et al., 2004). This includessimple datasets like a shoreline-database aswell as geological priors (e.g., fault information,seismo-tectonic regimes, etc.) and finally theexpert knowledge about generation models oftsunamigenic or tsunami earthquakes (e.g.,Bilek and Lay, 1999; Satake and Tanioka, 1999;Polet and Kanamori, 2000). Based on the priorinformation and the arriving evidence (seismicsource parameters) it is possible to automati-cally obtain a similar reasoning as an expertseismologist would perform, when trying tojudge on the soundness of data analysis as wellas the probability of strong ground shaking ortsunami generation and finally on the severityof possible consequences.There are two types of inputs into the system:1) detection messages of analysis results pro-vided by work packages 1 to 3 (i.e. epicentrallocation parameters, seismic moment esti-mates, source mechanism, rupture directionand speed, etc.); 2) prior information derivedfrom global seismicity data sets (Wadati-Benioffzone geometry, moment tensor solutions),geographical and geological data bases (e.g.digital elevation map and bathymetry, shorelinedatabase, tectonic boundary database). Out-puts are the automatically generated alert mes-sages. A message will be created whenever thealert level system changes its state based onthe evaluation of seismological evidence.

ReferencesAspinall, W.P., G. Woo, B. Voight and P.J.Bayter (2003). Evidence-based volcanology:

19

application to eruption crises, JVGR, 128,273–285.

Bilek, S. and T. Lay (1999). Rigidity variationswith depth along interplate megathrust faultsin subduction zones. Nature, 400, 443–446.

Boore, D.M. (2003). Simulation of groundmotion using the stochastic method, Pure andApplied Geophysics, 160, 635–676.

Campbell, K.W. (2003). Prediction of strongground motion using the hybrid empiricalmethod and its use in the development ofground-motion (attentuation) relations in East-ern North America, Bulletin of the Seismologi-cal Society of America, 93(3), 1012–1033.

Dahm, T. and F. Krüger (1999). Higher-degreemoment tensor inversion using far-field broad-band recordings I: Theory and evaluation ofthe method with application to the 1994Bolivia deep earthquake, GJI, 137, 35–50.

Douglas, J. (2004). Ground motion estimationequations 1964–2003 Reissue of ESEE Report01–1: »Á comprehensive worldwide summaryof strong-motion attenuation relationships forpeak ground acceleration and spectral ordi-nates (1969 to 2000)« with corrections andadditions (No. ESEE Report No. 04-0001-SM).London: Empirical College of Science, Technol-ogy and Medicine, Civiel Engineering Depart-ment.

Friederich, W. and J. Dalkolmo (1995). Com-plete synthetic seismograms for a sphericallysymmetric Earth by a numerical computationof the Green’s function in the frequencydomain. GJI, 122, 537–550.

Moreno, B., K. Atakan, K.A. Furlokken, S.Temel and O.J. Berland (2004). SAFE-Tools: Aweb-Based Application for Identification ofActive Faults. SRL, 75, 2, 205–213.

Polet, J., and H. Kanamori (2000). Shallowsubduction zone earthquakes and theirtsunamigenic potential, GJI, 142, 684–702.

Satake, K. and Y. Tanioka (1999). Sources ofTsunami and Tsunamigenic Earthquakes inSubduction Zones, PAGEOPH, 154, 467–483.

Scherbaum, F., F. Cotton, and P. Smit (2004).On the use of response spectral-reference datafor the selection of ground-motion models forseismic hazard analysis: the case of rockmotion, Bulletin of the Seismological Society ofAmerica, 94(6), 2164–2185.

Scherbaum, F., J.J. Bommer, H. Bungum, F.Cotton, and N. A. Abrahamson (2005). Com-posite ground-motion models and logic trees:methodology, sensitivities, and uncertainties,Bulletin of the Seismological Society of Ameri-ca, 95, 1575–1593.

Scherbaum, F., F. Cotton, and H. Staedtke(2006). The estimation of minimum-misfit sto-chastic models from empirical prediction equa-tions, Bulletin of the Seismological Society ofAmerica, 96(2), 427–445.

Schweitzer, J. (2001). HYPOSAT An enhancedroutine to locate seismic events. PAGEOPH,158, 277–289.

Wald, D.J., V. Quitoriano, et al. (1999). TriNet»ShakeMaps«: Rapid Generation of Instru-mental Ground Motion and Intensity Maps forEarthquakes in Southern California. Earth-quake Spectra, 15, 537–556.

Woo, G., and W. Aspinall, (2005). Need for arisk-informed tsunami alert system, Nature,433, 03 February 2005, 457–457.

20

WeraWarn: Real time detection of tsunamigenerated signatures in current mapsmeasured by the HF radar WERA to supportcoastal regions at risk

AbstractThe WeraWarn project aims at optimizing theWERA current and wave radar for the earlywarning of tsunamis, as well as to test the sys-tem’s capabilities and limitations in this con-text. To date, the WERA system has beendeployed to date to extensively measure oceansurface currents and wave heights of searegions covering several thousand square kilo-metres in real time and at high temporal andspatial resolution. These features will be usedfor detecting tsunami induced current signa-tures in the shelf break area. Thus a WERA sys-tem can produce warnings for the affectedcoastal regions. The investigation will be car-ried out in three steps: 1) Based on temporallyand spatially high resolution numerical simula-tions, the typical hydrodynamic properties ofan oncoming tsunami wave will be investigat-ed for various sea regions. This involves theevaluation of the influence of different shelfbreak geometries on the signatures in the sur-face currents. It is expected that the slope andorientation of the shelf break significantly con-tribute to the strength and propagation direc-tion of the observed signature. 2) The rela-tionship between typical tsunami current sig-natures, calculated with a radar backscattermodel with input from a model studydescribed above, and the actually measuredradar backscatter spectra serves as a basis forthe development of a specific tsunami early

warning algorithm. 3) In addition to theoreti-cal considerations and numerical simulations,a field experiment will be carried out in a thirdphase, during which a tidal bore of substantialheight, like it e.g. occurs in the HangzhouDelta in China, will be recorded with theWERA system. In this way, the detection capa-bility of the system can be determined.

1. IntroductionIn the aftermath of the natural disaster thatoccurred in December 2004 in the IndianOcean, the necessity for a reliable early warn-ing system has become the focus of attentionof politics, science and research. It has turnedout that existing technical solutions are notyet sufficiently reliable and thus hold the highrisk of producing false alarms. The aim of thisproject is to investigate the capabilities of afar range HF-Coastal Radar, which has beensuccessfully deployed in various locations du-ring many national and international fieldexperiments, to contribute to reducing thenumber of false alarms through the observa-tion of direct tsunami signatures. On the basisof theoretical considerations, numerical simu-lations and a final verification of the results byfield experiments, the possibilities and limita-tions of the WERA HF-Radar System and itssuitability for being a supportive or evenessential component of an early warning sys-tem, shall be investigated.

Stammer D. (1), Pohlmann T. (1), Gurgel K.-W. (1), Schlick T. (1), Helzel T. (2), Kniephoff M. (2)

(1) Institute of Oceanography, Center for Marine and Atmospheric Research, University of Hamburg, Bundesstr. 53,

D-20146 Hamburg

(2) Helzel Messtechnik Ltd., Carl-Benz-Str. 9, D-24568 Kaltenkirchen

21

2. Objectives and conceptThe development and testing of early warningsystems for minimizing the risk of damagecaused by tsunamis or cyclone surges com-prises several different stages. The proposedproject will make an outstanding contributionto the development and improvement of mon-itoring systems, to the development of modelsfor coupling remote sensing data and numer-ical simulations, as well as to the quality test-ing of early warning systems. Furthermore andfirstly, an estimate of the usefulness of an HF-Radar based early warning system will bemade for the most endangered coastal regionsworldwide. To this end, we will fall back onour own model calculations and on those ofthe AWI Project Group »Deutsche Finite Ele-mente Ozean Model«. The data and results ofthe project proposed here will be available forthe »New Spaceborne and Ground-basedMicrowave or UHF Systems for Detection ofTsunami« during the term of the project.

2.1. Development and improvementof monitoring systemsThe project WeraWarn will, by combining andadapting tested remote sensing technologyand numerical simulations, provide a new andpowerful component for a modern earlywarning system that fulfils the requirement ofproviding prompt, extensive and reliable infor-mation. The WERA Radar System, manufac-tured by the Helzel Messtechnik Companyand developed in collaboration with the Uni-versity of Hamburg (IFM/ZMAW), offers asimultaneous large area measurement ofocean currents, wave heights and wind direc-tion, i.e., desirable attributes of an earlywarning system »sensor.«To date, the HF-Radar has an up to 250 kmrange capability with a local resolution of upto 150 m and a temporal resolution down toless than 3 minutes. Thus a rapid detection ofsignificant changes is now already possible.Nevertheless, further technical system im-provements must be achieved in order torapidly detect the expected current signaturesthat a tsunami at the coastal shelf or a cycloneat sea would generate. In particular, the chal-

lenge is to continually analyse (filter) the con-tinuously generated current maps, or to bemore exact, the HF-Radar backscatter spectrathe maps are based on, in order to extract aset of parameters that is capable of reliablytriggering an alarm upon reaching a criticalconstellation.

2.2. Development and calibrationof a quantitative physical modelTo date, it has been possible to investigatehighly dynamic current processes in coastalregions with the HF-Radar. Tidal currents, drift-ing eddies, fronts and cyclone driven currents(Florida) have been recorded successively andused to drive numerical simulations for theforecasting of water movements (e.g., ship-ping route optimization, EUROROSE andWings-for-Ships EU projects). However, prelim-inary model generated current patterns of atsunami wave in the shelf area (e.g., Sey-chelles, Androsow) have to date never beenrecorded with an HF radar or a comparableremote sensing method, which possibly wouldallow a derivation of the topographicallyinvariant parameters of a tsunami wave.Due to the »fortunately«, in the statisticalsense, rare occurrence of tsunami events, itwill not be possible in the future to directlyascertain such parameters from the measure-ment data. Thus the modelling of the physicalprocess will be applied here in two ways tocompensate for the lack of measurement data.In a first step (cf. 4.3 Modelstudy Tsunami-Coast), appropriate models will be developedand operated to generate high resolution,small scale tsunami-similar coastal area currentpatterns which will be used in a second step(cf. 4.4 Modelstudy TsunamiHfSpec) to calcu-late the corresponding HF-Radar backscatterspectra. It will thus be possible to quantita-tively describe the chain of physical processesoccurring from the origin of the tsunami tothe HF-Radar spectrum. In addition, the modelresults will be used to conceptually supportthe planned field experiment (cf. 4.5 Fieldex-periment TidalBore).Subsequently, as data from the tidal boreexperiment becomes available, these data,

22

evaluated on the basis of experience gainedfrom the models, could possibly contributein an iterative process to improving themodels and/or optimizing the HF-Radar sys-tem characteristics.

3. A brief description of the basics

3.1. Modelling of tsunami wavesShortly after the occurrence of the disaster inthe Andaman Sea, Dr. Alexey Androsov carriedout model calculations at the ZMAW on thebasis of the model GNOM (General hydrostat-ic/non-hydrostatic Ocean Model).The simulation provided surface displacementsand current velocities for the area of sea infront of Phuket and the Seychelles as it isshown for the Seychelles in Fig. 1. The modelcovered an area that included the islands ofMahé and La Digue, the shallow shelf and thedeep sea region bordering the shelf.The simulation was based on the assumptionthat the first quake occurred near the NikobarIslands and triggers a tsunami. The tsunamiwave propagates at a high velocity towardsthe Seychelles and reaches the northeastmodel boundary with a wave height of 50 cmat model time t = 0. From this point on, thehigh energy wave is influenced more andmore by the structure of the shelf. The model

simulation confirms the assumption that at adistance of 150 to 200 km from the island ofLa Digue, significant alterations (changes inmagnitude and direction) of the current fieldoccur along the shelf break (edge) at modeltime t = 72 min, cf. Fig. 1.The extreme alterations of the current fieldand the pronounced orientation to the iso-baths could serve as indicator for an oncom-ing tsunami. From this point on, it takesanother 100 minutes until the tsunami reach-es the coast of La Dique, in agreement withknown propagation times. After another30 minutes, the tsunami arrives at the largerisland of Mahé. Velocities of up to 2 m/s areobserved in the current field, which is an indi-cation that with high probability, the associat-ed current patterns could be detected andtracked by an HF-Radar. The radar integrationtime must only be short enough (ca. 4 to 5 min-utes) to resolve the dynamics of the event.

3.2. Theory and numerical simulationA Tsunami generally is generated by an abruptvertical displacement of a large amount ofwater in the ocean. This can be induced dueto earthquakes, volcanic eruptions or land-slides. Also meteorite impacts carry the poten-tial to trigger a Tsunami. For example onDecember 26th in 2004 we were faced with

Figure 1: Simulated current signa-tures at the shelf edge of theSeychelles due to a propagatingtsunami

23

the worst effects of those kind of waves in theAndaman Sea. On a wide scale from 500 to1000 km, the ocean floor has been displacedby an earthquake, that itself was possibly gen-erated by another earthquake on the otherend of the Indic-Australian-Plate near Antarc-tica two days before. Due to the imbalance(you can compare it with a plate of ice in thearctic ocean) the earthquake in the AndamanSea raised with a vertical displacement of thesea bottom by 10 m to 30 m.But other natural events are able to induce atsunami as well. For example, the enormousmix of gas and steam which is ejected duringa submarine volcanic eruption may generate atsunami wave due to the big displacement ofthe water column.

Figure 2: Phase velocity as a function of the water depth Figure 3: Current velocity as a function of the waterdepth

In theory a Tsunami wave is a long wavebecause the relation between the wavelengthof the tsunami (~100 km) and the waterdepth (4 to 5 km) is very large. Those wavesare called shallow water waves and they aremostly influenced by the bottom topography.The phase velocity c can be characterized byc = sqrt(g*h), cf. Fig. 2.The total energy of a tsunami in the deep seais nearly balanced, which leads to the effect ofsmall amplitudes (0,1 to 1 m) and very highphase velocities of this wave in the area of theepicentre (nearly 900 km/h). But when thewave is reaching islands or shelf edges thewater depth is decreasing (that means thevelocity is also decreasing) and the potentialenergy is extremely increasing (so we get a

Figure 3.1: Bathymetry of the Seychelles with transect at4,9° South

Figure 3.2: Variance of currents along this section in amoving window of size 5 km

24

much higher amplitude of about up to 30 m)contrary to the velocity of propagation ofkinetic energy, that is decreasing (30 km/s).Fig. 2 illustrates the decrease of the phasevelocity due to the decreasing water depthfrom 700–800 km/h in the deep sea down to100–150 km/h, while the wave is reaching thecontinental shelf edge (100 to 200 m waterdepth). These effects are accompanied byincreasing amplitudes and current velocities.Fig. 3 shows the relation between currentvelocity and water depth for a wave propa-gating from deep to shallow water. For a realbathymetry, like it is shown in Fig. 3.1 for theSeychelles islands, at some points currentvelocities of about 1 m/s and more can beobserved.Supposed we would acquire the surface cur-rents along the section above by the usage ofa radar system and by determining the vari-ance of the current velocity in radar cells withan extent of about 5 km, we will get signifi-cant signal peaks in variance (cf. Fig. 2.2) atthe continental shelf edge, which in the endcan be used as an indication for a Tsunamiwave.Exact statements concerning the spatial andtemporal behaviour of signatures in the veloc-ity field, which are related to tsunamis, needmodel studies with a very high resolution inthe order of several hundred metres. Accord-ing to theory tsunami, waves can be consid-ered as »long waves«. They produce purebarotropic (vertically homogenous) currents asa first order approximation. Thus for the pro-posed investigations it is sufficient to apply atwo-dimensional model. Moreover, the baro-clinic (density induced) part of the circulationcan be neglected since for the wave motion itonly plays a minor role. Therefore a simpletidal model can be applied in the frame of theinvestigations. Such type of model was alreadyused at the Institute of Oceanography in Ham-burg in the seventies. Due to the enormousdevelopment in computer technology it isnowadays possible to use these relatively sim-ple water elevation-current model with anextremely high spatial resolution, which allowsto implement the above mentioned grid dis-

Figure 4: Typical HF radar backscatter spectrum

tances of several hundred metres. The largestproblem, if scenarios of actual region shouldbe simulated, is the lack of sufficiently high-resolution topographic data. In certain cases itmay be necessary to carry out side-scan-sonarinvestigations in the respective section of theshelf break.

3.3. Basics of HF radar technicsThe use of HF radar in oceanography hasbeen initiated by Crombie (1955) and has beenapplied for the measurement of ocean currentmaps at the Institute of Oceanogra-phy, University of Hamburg, Germany, since1980 within the scope of many field experi-ments1. Due to progress in system hardwareand processing software driven by the oceano-graphic requirements, it is now also possible tomeasure ocean waves and wind direction atan increased spatial and temporal resolution.The basic physics utilized by an HF radarinstalled at the coast, which is ground wavepropagation of electromagnetic decameterwaves and Bragg scattering at the ocean sur-face, allow the observation of processes farbehind the horizon (up to 250 km).The electromagnetic waves are transmitted atthe coast and propagate along the ocean sur-face. A small fraction of the transmittedenergy is scattered back due to the rough-ness of the ocean surface and is received bythe radar. After processing of the echoes,radar backscatter spectra (cf. Fig. 4) are avail-

25

able. Based on these spectra, surface current,wave height and wind direction can beprocessed. The most important parts of thespectra are the ›First Order Peaks‹ and the›Second Order Returns‹.In case of a propagating tsunami wave, specialcurrent signatures can be expected at theocean surface, which influence the shape ofthe backscatter Doppler spectrum. Numericalsimulations and experiments will be used toinvestigate the influence of the tsunami on theDoppler spectrum. Later on, these findings canbe used to deduct information on a tsunamiwave from a Doppler spectrum.As a single HF radar system measures the radi-al component of surface current, towards oraway from the radar, it is necessary to installtwo sites at different locations along the coast.The two radial components can be combinedto give the two-dimensional current vectors.Using this set-up, it is possible to measureactual current maps covering large areas. Fastchanging features in the current field, e.g.drifting eddies, fronts, or tsunami waves, canbe monitored and resolved, because the sys-tem is able to measure actual current mapswithin minutes. Fig. 5 shows a sequence ofcurrent maps measured at the French coastnear Brest with a temporal resolution of12 minutes.Successful tsunami detection requires thatactual current maps are available at shorttime intervals, which must be processed inreal time to release a tsunami warning asearly as possible.

3.4. Basics and requirements in hydrodynamicnumerical simulationsThe Institute of Oceanography of HamburgUniversity has decisively participated in thedevelopment of numerical hydrodynamicalmodelling since the mid fifties (Hansen, 1952,Sündermann, 1966). Until the mid seventiestwo-dimensional barotropic, wind and tidal

Figure 5: Varying currents in 12 minute intervals, typical currentamplitudes are in the range of tsunami induced amplitudes

26

driven models have been in the focus ofresearch activities (Maier-Reimer, 1977). Sub-sequently, a development towards three-dimensional, baroclinic models was observed(Backhaus, 1979; Maier-Reimer, 1979). At thebeginning of the eighties IfM was the firstEuropean institute which conducted long-termsimulations for actual periods. Within this typeof modelling, the circulation model HAMSOMplayed a key role. Due to the semi-implicit for-mulation of the gravitational waves and thevertical exchange, it was possible for the firsttime to perform simulation for periods ofdecades using realistic forcing conditions(Backhaus and Hainbucher, 1987). ThomasPohlmann was one of the leading personsresponsible for the development of a prog-nostic baroclinic North Sea model. This in par-ticular concerns the treatment of the thermalstratification (Pohlmann, 1996, 1997).Furthermore, Th. Pohlmann participated in anumber of national (e.g. ZISCH, PRISMA) andEuropean projects (e.g. NOMADS2 (Delhez,2004)), several times in a leading position asprinciple investigator. In these projects thehydrodynamical model has been applied suc-cessfully. In particular in the BMBF-project»Storm Surge Development« the foundationsfor the applied tsunami project have been laid.In this project a prognosis for the developmentof storm surges along the North Sea coast hasbeen generated taking into account future cli-mate development. For this purpose climatescenarios calculated by the Max-Planck-Insti-tute for Meteorology have been facilitated.Presently Th. Pohlmann is principle investigatorin a GLOBEC-D subproject, and in the projectsVietnam Upwelling and Siak River Dispersion.In all of these projects, the hydrodynamicalmodelling plays a dominant role.

4. Description of the working plan

4.1. Basic structure of the concerted projectThe cooperation between the »Zentrum fürMarine und Atmosphärische Wissenschaften(ZMAW) and Helzel Messtechnik (HMT)« isaimed to allow the development of an opera-

tive component of a tsunami early warningsystem (TEWS). The extraordinary experienceof HMT in the field of hardware design andsystem technics, especially in radar technics,will unburden ZMAW from hardware relatedwork and therefore allow to concentrate onthe data analysis and development of numer-ical models including tsunami-wave propaga-tion and radio backscatter simulations.For the WeraWarn project, 4 tasks can be dis-tinguished, cf. Fig. 8 (project overview), whichwill be discussed in the following sections.

4.2. Model study »TsunamiGlobal«In TsunamiGlobal, an existing storm surgemodel will be applied to the world ocean. Theplanned resolution will be 10′ (approx. 20 km).Due to the implicit solution scheme for theexternal gravity waves no limitation of thetime step must be considered. The time stepcan be chosen purely due to physical require-ments. With the global model, qualitativestudies with respect to the tsunami risk will beconducted for different coastal zones. Thisanalysis will mainly be of comparative nature,i.e. the risk potential of different coast zoneswill be compared. The tsunami will be trig-gered at region with known tectonic instabili-ties. Also the duration and strength of the per-turbation will be defined by means of historictsunami events

4.3. Model study »TsunamiCoast«In TsunamiCoast, the storm surge modeldescribed above will be applied to selectedcoastal regions. The resolution will bebetween 0.25’ (500 m) and 1’ (approx. 2 km)depending on the region of interest. For thisstudy the parallelized version of the modelwill be employed in order to obtain an opti-mal performance on the vector computerat the »Deutsches Klima Rechenzentrum«(DKRZ). The parallelization will be performedby means of the domain-splitting scheme.The vectorization of the implicit part of themodel code is carried out with help of thered-black-algorithm, which is known to showvery good numerical properties for diagonaldominant matrices.

27

At the beginning, the model will be validatedby means of historical tsunami events. In thiscontext, it is possible to calibrate relevantparameters like the bottom friction coefficient.Subsequently, scenario runs for selectedtopographies will be conducted. The resultingsignatures of the surface velocities will beanalysed in combination with the governingparameters like shelf width, topography gradi-ent, distance of epicentre and strength andform of the perturbation. Finally, the signa-tures and the derived analyses will be madeavailable to the other project partners. Theywill use this information to more accuratelydescribe the influence of tsunami waves onthe sea surface currents, which is the param-eter that is finally detected by the HF radar.

4.4. Numerical simulation »TsunamiHfSpec«The ocean surface current data simulated bythe high resolution tsunami model are avail-able at a resolution of about 100 m in spaceand 10 s in time. Together with ocean wavespectra, which are derived from the climatol-ogy, they are passed to a radar backscattermodel, which calculates the highly variablebackscatter Doppler spectra across the shelfedge for the radar resolution cells as a func-tion of time. This backscatter model is used to

investigate the characteristic structure of atsunami signature, i.e. how long can it be seenwithin a resolution cell, what is the variance ofthe Bragg frequency, how fast does the signa-ture move through the resolution cells, andhow do these characteristics depend on thetopography. The results of this investigationwill be used to optimize the required spatialand temporal resolution of the radar.In a next step, typical design parameters for aspecial filter that allows to recognize tsunamisignatures in HF radar measurements will bederived from EM backscatter simulations. Thesensitivity and response characteristic of sucha kind of filter can be tested by superimpos-ing a simulated tsunami signature to real HFradar data.On this basis, the fast propagating signatureshould be observable and assessable withinthe normal ocean circulation dynamics gener-ated by tides, fronts and wind influence.

4.5. Field expriment »TidalBore«Beside of all the theoretical and numericalinvestigations, the capabilities of HF radar todetect tsunami waves should be tested in afield experiment. For this purpose the radarsystem must be adapted and optimized to thespecial task, which is mainly in the responsi-

Figure 6: Area of observation with 3 radar coverage at Hangzhou delta, China

28

bility of Helzel Messtechnik. The followingparagraphs summarise the main aspects of thesystem optimization:– The excellent signal to noise characteristic of

a WERA system allows for modifications likethis without the risk to reduce the reliability.The higher temporal resolution can be achiev-ed by means of optimised signal processingalgorithms and more powerful computers.

– The required higher spatial resolution can beachieved by increasing the operation band-width of the radar. Since our radar has a bro-adband concept, this higher resolution istechnically possible, but for long rangesystems the generated data volume wouldbe hard to handle. This problem is again amatter of signal processing optimisation andadaptation to the individual application (areaof interest).

– As a component in a national disastermanagement system the radar needs to bevery reliable. Actually the rate of availabilityof data from the entire field of view is about90%. The requirement for a sensor compo-nent of a disaster management system isstronger, almost 100% should be goal. Thereason to loose data is typically external elec-tromagnetic interference (EMI). This problemcan be strongly reduced by using more thanone frequency band for the radar and usinga multi-frequency system that automaticallyselects the best frequency band. Even if theWERA already is designed as a broadbandsystem, for this purpose special antennas,input filters and multiplexers as well as soft-ware need to be developed.

Our main goal is to carry out a field experi-ment where we may test the radar system,

Figure 7: Tidal bore at Hangzhou delta, China

software, warning strategy, and the alarmchain under »tsunami conditions«. The vari-ability of the current pattern for a tsunamiwave in shallow shelf water, where it gains inheight and runs with reduced speed towardsthe coast, should at least be comparable tothat of a tidal bore cf. Fig 7. A practical testshall determine if these assumptions hold truein nature.There are, however, no sufficiently pro-nounced tidal bores in Europe, that occur soregularly that there is a high probability toobserve one during a measurement campaign.Searches to date have produced three possiblelocations:1. Canada, Bay of Fundy or PETITCODIAC

(Amplitude of the Bore ca. 1m)2. Brasil, Amzonas Delta or ARAGUARI

(Amplitude ca. 4–6 m)3. China, Hangzhou Delta Bore QUIANTANG

DRAGON (Amplitude ca. 7–9m)The last location No. 3 (China, cf. Fig. 6) is pre-ferred by us, since the tidal bore here is mostpronounced and predictable. Second and thirdchoices would be location No. 2 (Brazil) andlocation No. 1 (Canada) respectively.

5. ConclusionThree independent research institutes (UniHamburg, Uni Sheffield and the researchgroup of Codar Ocean Sensors) came to thesame conclusion, namely that an approachingtsunami will generate a significant ocean cur-rent signature when it reaches the continentalshelf. Due to the dramatic reduction of thetsunami wave’s velocity in shallow water, therecan be enough time for a pre-warning if theshelf edge is more than 50 km offshore. Thedetection and analysis of this effect can pro-vide a reliable forecast of the effects at specif-ic coastlines or harbour areas. Since evensmaller tsunamis can cause loss of lives ordamage of coastal properties, such a localforecast is important. It is well known thatwarnings will be effective only if the numberof false alarms are minimal. A local detectionsystem can increase the reliability of a nation-al tsunami warning system.

29

6. Sketch of project overview

Figure 8: Combination of working modules

30

ReferencesBACKHAUS, J. O., 1979: First results of athree-dimensional model on the dynamics inthe German Bight. In: Proc. of 10th Internat.Liége Colloq. on Ocean Hydrodynam. Ed.: J. C.J. Nihoul. Amsterdam: Elsevier, ElsevierOceanography Series 25, 333–349.

BACKHAUS, J. O. UND D. HAINBUCHER,1987: A finite difference general circulationmodel for shelf seas and its application to lowfrequency variability on the North EuropeanShelf. In: Three dimensional models of marineand estuarine dynamics. Ed.: J. C. J. Nihoul undB. M. Delhez, E., J.M, P. DAMM, E. DE GOEDE,J.M. DE KOK, F. DUMAS, J.E. JONES, J. OZER,T. POHLMANN, P.S. RASCH, M. SKOGEN,R. PROCTOR (2004): What can we expect fromshelf seas models: the NOMADS2 Project. Jour-nal of Marine Systems, 45, 39–53.

HAINBUCHER, D., J.O. BACKHAUS, 1999: Cir-culation of the eastern North Atlantic andnorth-west European continental shelf – ahydrodynamic modelling study. FisheriesOceanography, 8, Suppl. 1, 1–12.

HANSEN, W., 1956: Theorie zur Errechnungdes Wasserstandes und der Strömungen nebstAnwendungen. Tellus, 8, 287–300.

MAIER-REIMER, E., 1979: Some effects of theAtlantic circulation and of river discharges onthe residual circulation of the North Sea. Dt.Hydrogr. Z., 32, 126–130.

POHLMANN, T. (1996): Predicting the Thermo-cline in a Circulation Model of the NorthSea–Part I: Model Description, Calibration andVerification. Continental Shelf Research, 16/2,131–146.

POHLMANN, T. (1997): Estimating the Influ-ence of Advection During FLEX’76 by Means ofa Three-Dimensional Shelf Sea CirculationModel. Deutsche Hydrographische Zeitschrift49 Vol. 2–3, 215–225.

POHLMANN, T. (2005): A meso-scale model ofthe central and southern North Sea: conse-quences of an improved resolution, acceptedby Continental Shelf Research.

SÜNDERMANN, J. 1966: Ein Vergleich zwi-schen der analytischen und der numerischenBerechnung winderzeugter Strömungen undWasserstände in einem Modellmeer mitAnwendungen auf die Nordsee. Mitt. Inst.Meeresk. Univ. Hamburg Nr. 4, 77 S.

CROMBIE, D. D. (1955): Doppler spectrum ofsea echo at 13.56 Mc/s., Nature 175 (4449):681–682

K.-W. GURGEL, H. H. ESSEN, and T. SCHLICK,2003: The use of HF radar networks withinoperational forecasting systems of coastalregions. in ›Building the European Capacity inOperational Oceanography‹ 3rd InternationalEuroGOOS Conference, Pro-ceedingspp. 245–250, Elsevier, ISBN 0 444 1550 X.

K.-W. GURGEL, 1999: Applications of CoastalRadars for Monitoring the Coastal Zone. Pro-ceedings EUROMAR Workshop’99. Publishedby: EUROMAR Office, Agência de Inovação,S.A., Av. dos Combatentes, 43, 10°C/D, 1600-042 Lisboa, Portugal, pp. 21–30.

K.-W. GURGEL, H.-H. ESSEN, and T. SCHLICK,2000: Eddy dynamics off the Norwegian coast,recorded by HF radar. IGARSS’2000 Confer-ence, Proceedings, pp. 1839–1841.

1 http://ifmaxp1.ifm.zmaw.de/Experiments.shtml

31

Early Warning System for Transport Lines(EWS TRANSPORT)

AbstractBased on recent improvements of early detec-tion methods for earthquakes as well asadvances in the field of communication andinformation technologies, new risk minimiza-tion strategies for railbound transportationsystems are developed.

1. Goals of the project

The purpose of the proposed research projectis to develop and test an early earthquakewarning system that minimizes the risk ofdamage for transport lines. The term »trans-port lines« is meant to comprise transportoperation, vehicles (including passengers andgoods), and infrastructure.The development and testing of an earlywarning system for transportation lines willbe carried out first for railbound transporta-tion systems. On one hand, this transporta-tion mode is particularly vulnerable and riskminimization measures are more difficult toimplement, e.g. because of the long brakingdistance of trains. On the other hand, inrailbound traffic, every train movement isplanned in advance and controlled from theoutside, and thus characterized by very highstandards of organization, control, and safety.Therefore, this mode of transportation is pre-destinated for applying specific and coordi-nated emergency measures from outside, andhence well suited for testing the efficiency ofthe proposed early warning system. The expe-rience gained here will be useful in the devel-

opment of information systems for othertransport lines.For many transportation systems there existtraffic control centers, which can interfere andinfluence the flow of traffic more of lessdirectly. In the case of an earthquake, trafficcontrol centers will decide whether and wherein a particular region, traffic is prevented toenter an endangered bridge or tunnel. Fur-thermore, an immediate speed reduction ofthe vehicles in that region may be initiated. Inthe case of railbound transportation systems,passenger safety requires that certain opera-tion situations, e.g. trains stopping on a bridgeand train encounters (danger of collapse, dan-ger of panic and ensuing damages) are avoid-ed. In railbound traffic, vehicle movement can-not only be influenced by the train driver butalso by an external train control center. Thismeans that one has enhanced possibilities ofinterfering from the outside, but also entailsadditional decision making problems. After anearthquake, the train control center mustdecide which parts of the network are acces-sible. In general, the train driver cannot reactsufficiently quickly even if an obstacle on thetrack (e.g. due to a land slide) or track buck-ling is recognized ahead of time, except whendriving in the low safety mode »running atsight« (vmax ≤ 40 km/h).Figure 1 and the following sections provide anoverview over the three research and develop-ment areas of the proposed early warning sys-tem. In chapter 4 the methods used in eacharea are discussed in more detail.

Wenzel F. (1), Kühler T. (1), Hohnecker E. (2), Buchmann A. (2), Schedel F. (2),

Schöbinger F. (2), Bonn G. (3), Hilbring D. (3), Quante F. (3)

(1) Geophysical Institute, University of Karlsruhe

(2) Department of Railway Systems, University of Karlsruhe

(3) Fraunhofer Institute for Information and Data Processing, Karlsruhe

32

(I) New early detection methods are used todetermine important parameters such as mag-nitude, location, time, and spatial distributionof an earthquake. The seismic hazard for a par-ticular region is then projected onto transportlines within the range of a given control center.With the proposed early detection methods,the necessary information can be obtainedwithin a few seconds, and depending on theepicentral distance warning times up to1 minute or more are achievable.(II) For risk minimization detailed emergencyplans are carried out by various traffic controlcenters, based on preconceived hazard anddamage catalogues from which the appropria-te measures can be chosen. As mentionedbefore, a typical action would be the stoppingof a train before it enters a damaged part ofthe network. Furthermore, EWS Transport willproduce an early estimate of the damage in thenetwork that has been caused by the earth-quake. Finally, the proposed standardized earlywarning system will allow a continuous moni-toring of the network state also when there isno imminent danger.(III) In view of the short warning times and thelarge number of alternatives, the use of new

information and communication technologies(IaC) is absolutely necessary in order to accele-rate the exchange of information and to sup-port the decision making processes of theresponsible managers. Using so called »messa-ge brokers« data transfer between quite differ-ent systems is facilitated. In addition to con-ventional radio communication, the railwayspecific GSM-R standard is suitable for a priori-zed and fast signal transmission, in order toexecute the automatic emergency proceduresand to exchange information between thoseaffected by the earthquake. The use of geo-information system (GIS) standards facilitatesthe creation of shake maps, and high levelarchitecture (HLA) compliant IaC systems allowthe efficient use and coupling of differentmodels needed for hazard prognosis simula-tions. The latter are an essential building blockof the decision support system.The collaboration between the three differentresearch and development areas (see Fig. 1)can be characterized as follows: experts fromthe fields »early detection« and »risk mini-mization« develop and compile the requiredscientific and practical knowledge. This com-prises the development of early detection

Figure 1: The three research and development areas in EWS TransportEarly detection (I) is performed by seismological basis stations, where relevant earthquake parameters are derived fromthe raw data. The former are then transmitted via radio communication to various traffic control centers where a riskminimization procedure (II) is started by choosing appropriate emergency plans and measures of which early warning isone of the most important. Efficient IaC technologies (III) are necessary to support the decision making process becauseof the short warning times and the complexity of the problem.

33

methods, and the compilation of hazard anddamage catalogues for estimating the immi-nent danger in transport operation and thedamage done on various infrastructure com-ponents. On the basis of these catalogues anda given set of earthquake parameters, appro-priate emergency plans will be automaticallyselected and executed. In the IaC field, com-puter scienticists develop the functionalitiesthat allow a simulation of the various steps inthe early warning chain. The results of thesesimulations can then be used in decision sup-port processes. The structure and criteria thatunderly the decision support system aredefined by the experts from the areas »earlydetection« and »risk minimization«.The proposed research and development iscarried out by an interdisciplinary team ofresearchers and practitioners from universities,research institutes, international engineeringfirms and industry.

2. Relevance to societyAs recent events have shown, world commu-nity is no longer willing to accept the catas-trophes caused by major earthquakes as merefate. The motivation for the proposed projectis based on the knowlegde that majoradvances in the fields of sensor-, computer-,and IaC technologies have been made. Thisopens up new possibilities for minimizing thedamage through the use of an early warningsystem in connection with support systems fordecision making.In recent years, through a worldwide effortnumerous projects have been started, whichaim at risk reduction and an improvement ofdisaster management in the event of an earth-quake. In Europe the main emphasis has beenput on projects, which aim at a reduction ofthe risks accompanying an earthquakethrough preventive measures and at improvingthe catastrophe management. In this contextthe following projects are noteworthy:– ORCHESTRA (http://www.eu-orchestra.org):

aims at an open architecture for riskmanagement with emphasis on prevention(with participation of IITB, see below)

– OASIS: catastrophe management system(http://www.oasis-fp6.org/)

– WIN (http://www.win-eu.org): IT system,which provides a connection to technologi-cal systems, which are useful for damageprevention and in the case of damage

– LESSLOSS (www.lessloss.org): a coordinatedeffort of 46 European partners for earthqua-ke hazard prognosis, assessment of itseffects on the environment, such as citiesand infrastructures, and disaster prepared-ness and protection strategies

The present project incorporates pertinentresults from these EU projects.

3. Current state of the art

3.1. Early warning systemAn early warning system is usually character-ized by a network of detectors, e.g. accelero-meters, which are uniformly distributed overan earthquake prone area so that the distancebetween an assumed epicenter and the near-est detector is minimized. The seismic wavesemitted by an earthquake can be classified inbody waves (compressional and shear waves)and surface waves. The shear (S-waves) andsurface waves have the largest amplitudes andcause the greatest damage. However, theirpropagation velocity is only about half of theP-wave velocity. When fast P-waves arrive atthe nearest detector, these data are electro-magnetically transmitted to a central informa-tion processing facility. There, the raw detec-tor data are converted into a few significantparameters, such as magnitude, location ofthe epicentre, and beginning of the earth-quake, including an estimate of their reliabili-ty. These parameters are then transmitted byradio communication to potential users, e.g.the train control centers in the case of rail-bound transportation.The basic idea of an early warning system is totake advantage of the time difference betweenthe arrival of a radio signal propagating withthe speed of light vem ~ 300000 km/s and seis-mic shear or surface waves, which propagatewith a typical velocity of vS ~ 3 km/s. For exam-ple, assuming that a vulnerable structure is

34

located at a hypercentral distance of 60 km,this leads to a time difference of 20 secondsbefore the S- and surface waves reach thisstructure. This time difference has to be usedby intelligent information systems in order tominimize the damage.Applications of early warning systems existonly in few places and only in simple form,e.g. in Japan (NAKAMURA 2004), Taiwan (WU,1998, 1999; TSAI, 1997), Turkey (ERDIK, 2003),Mexico (ESPINOSA, 1995), California (ALLEN,2003). Either the signals of different seis-mometer stations are transmitted to a centralprocessing station or they are processed direct-ly at the site of the individual stations. Usual-ly, only a few seismological parameters areextracted from the raw data. Threshold crite-ria are used as alarm signals. With presentearly warning systems, important informationrelevant for hazard prognosis and damage pre-vention, such as hypocenter location, propa-gation direction of the seismic waves, andshake maps is rarely used.However, in particular for railbound systemswith their narrow tolerances of track align-ment parameters, more detailed informationconcerning the propagation direction andamplitudes of seismic waves in the three spa-tial dimensions could be essential. This infor-mation could be used in assessing the impactof an earthquake on train operation and vari-ous infrastructure elements with higher accu-racy and for choosing the most suitable emer-gency measures.

3.2. Risk minimization for transport linesCurrently, there are no early warning systemapplications that focus on the special require-ments of transport lines, with the exception ofa Japanese system for railways that has beendeveloped since the 1980s. According toNAKAMURA the UrEDAS (Urgent EarthquakeDetection and Alarm System) for railboundtransportation is worldwide the only earlywarning system for transport lines in operation(NAKAMURA, 2004). In recent years it has beenrepeatedly activated. In case of an earthquake,single stations equipped with 3D accelerome-ters detect the P-waves, calculate its magni-

tude, and within 4 seconds transmit a radiowarning signal to endangered sites within arange of about 200 km. A standard emer-gency measure is then, for example, the deac-celeration and stopping of a fast Shinkansentrain. Each single UrEDAS station is able tomeasure, calculate, and transmit correspond-ing warning signals. A higher level networkorganization is not used.Possible disadvantages of the system arethe ones mentioned in section 3.1, and therelatively low network density. A higher den-sity network of P-wave detectors could helpin saving valuable seconds, and in provid-ing a more detailed and more reliable dam-age prognosis.

3.3. IaC technologiesIn seismology, complex IaC technologies, e.g.,for data fusion, have so far only been used ininvestigations of recorded signals after anearthquake. Recent progress in detector andcomputer hardware accompanied by simulta-neous cost reduction, in connection withadvances in the fields of early detection meth-ods, model building, simulation, and decisionsupport systems, suggest that an applicationin early warning systems and risk minimizationprocedures is promising.

4. Employed methods

4.1. Early warning methodologyGround motion caused by an earthquake con-sists of a wave train on three components (ver-tical and horizontal) with a length thatdepends on the magnitude and ranges from10 seconds (M = 5) to a minute (M = 8). Theacceleration signal starts weakly with com-pressional waves arriving first and increaseswith time to a damaging level associated withshear waves or surface waves.For a given magnitude, hypocenter, andsource time, ground motion can be estimatedat any location (a) by stochastic modelling, or(b) by empirically determined attenuation rela-tions of various parameters, such as PeakGround Acceleration (PGA). Traditionally,source parameters are determined with a seis-

35

mological net of stations. Although this pro-vides reliable values it takes time (at least oneminute) and relies on functioning communica-tion between stations and a data center (TENG,1997). Alternatively new methods have beendeveloped in the past years that utilize typi-cally the first 3 seconds of the weak initialarrivals of a single station in order to predictthe magnitude. A method developed at theGeophysical Institute in Karlsruhe bridges thegap between single station alarm systems andthose requiring records from an entire networkwith the Artificial Neural Network methodolo-gy (BÖSE et al., 2004, 2005). It allows to issuean alarm with one station, but upgrade it witheach additional available record continuously.This approach should be extended to estimatenot only ground motion at one or more spec-ified sites but also to determine source param-eters and relative amplitudes of groundmotion components. The feasibility of compo-nent-specific prediction of ground motionamplitudes is assesed by analysis of existingearthquake data bases (e.g. data from theJapanese K-net and Kik-net).Basically the following information can beextracted from seismic signals recorded in thevicinity of railway lines:

– After 3 seconds:Has an earthquake with damage potentialoccurred? What is its magnitude?

– After a few more seconds:When will the strong motion phase arrive atthe railway line in which location? How isground motion distributed on line-parallel,line-perpendicular, and vertical compo-nents?

– After minute(s):Fairly accurate map of ground motion alongthe railway line (shake map), that allows toassess the state of the line and potentialdamage. This requires detailed knowledgeof geotechnical parameters along the rail-way line so that site effects and secondaryphenomena such as liquefaction and landsli-de potential are included. The geotechnicalparameters are usually known along railwaysas they are required for its structural design.

Assessment of the precision and reliability ofseismic prognoses in short time with ArtificialNeural Nets represents a major challenge forthe seismological part of this project, particu-larly with regard to the noisy environment inthe vicinity of railway lines. Hence, seismicmeasurements are carried out along traintracks (see Fig. 2) to obtain the required noise

Figure 2: Seismogram showing 45 minutes of ground velocity recorded by a seismometer in 4.83 m distance from thetrain track. The upper plot shows the component of ground motion oriented perpendicular to the train track (E), themiddle plot the component parallel to the track (N) and the lower plot the vertical component. The periods with dis-tinct high amplitudes are caused by passing trains.

36

model. The model will allow an appropriateevaluation of the possibility to extract infor-mation on earthquake ground motion fromaccelerometers along railway lines. Further-more, the measurements provide a basis forthe development of methods for continuoustrack state monitoring.

4.2. Development of strategiesfor risk minimizationA risk minimization procedure will be suc-cessful if the relations between a given three-dimensional signal strength and its conse-quences for transport operation and trans-port infrastructure are known in some detail.This knowledge is available in the form ofphysics and civil engineering reference man-uals, and in the form of special models (e.g.derailment models in railbound systems). Anintegrated application of this knowledgecombined with the information about thespatial and temporal distribution of strongground motion (shake maps) leads toimproved hazard estimates and impact prog-nosis. The latter serves as a basis for applyingproper emergency measures in the regionaffected by an earthquake.This process may be considerably acceleratedif one can rely on pre-existing earthquake haz-ard and damage catalogues containing variousimpact scenarios as well as correspondingemergency plans and measures. The choice ofappropriate emergency measures from thesecatalogues can be supported by a computer ifthe aforementioned relations have been for-malized as an expert system.In the case of an earthquake with damagepotential the responsible emergency managersmust make numerous decisions under ratherdifficult circumstances:– There are numerous alternatives– The preference for a particular alternative is

not clear– There are a number of criteria that have to

be taken into account– The required information is not sufficiently

accurate– Decisions have to be made under an enor-

mous time pressure

The complexity of the problem suggests thatonly a consequent formalization of the under-lying knowledge of hazard and damage esti-mates on one hand, and of suitable measuresfor risk minimization, on the other hand, via acomputer is able to support the responsibleemergency managers.

4.2.1. Compilation of hazard, damage,and emergency measure cataloguesIn Europe railway traffic is sometimes orga-nized and supervised by few national traincontrol centers (e.g. ~5 in Germany). Thesetrain control centers can interfere and affectindividual trains over a large area if a corre-sponding warning from a superior agency(e.g., meteorological service, Department ofthe Interior) has been issued. For example,train control centers can stop and redirecttrains within their range and coordinate emer-gency measures.If track alignment is perturbed by an earth-quake the probability of derailment increasesrapidly with increasing speed, in particular incurves. Even a moderate speed reduction thathas been initiated by an early warning systemimplemented in train control centers is able toconsiderably reduce the severity of an acci-dent, or possibly avoid an accident altogether.Therefore, one goal of EWS Transport is thecompilation of relevant catalogues containingvarious hazard, damage, and emergency planscenarios based on a given earthquake sce-nario. For this, the physical properties of trackinfrastructure and engineering works, thedynamics of train motion, and the train posi-tions and velocities on the network have to beknown. To achieve this goal we can rely on theexperience and expertise of a network ofresearchers and practitioners from seismology,civil engineering, train operation, and infor-mation science.

4.2.2. Track network as an earthquakedetector and infrastructure monitorAn interesting aspect of railbound systems isthe principal possibility to employ the existingrail network with its internal sensors and elec-trical wiring for early detection and for infor-

37

mation transmission. The earthquake earlywarning system would then be part of thestandard train control and protection system.Because the train control and protection infra-structure is currently redesigned with the aimof achieving technical interoperability betweendifferent EU member states (BERGER, 2004), itseems timely to ask the question:How must the train control infrastructurebe modified so that it can be used forearly detection purposes, fast damage mapgeneration, as well as for continuous trackstate monitoring?To answer this question, a feasibility study isperformed and suitable sensors and IaC solu-tions are developed during the first phase ofEWS Transport. In a second phase, a testtrack section will be designed and build(demonstrator) in order to test these compo-nents of an early warning and monitoringsystem in practice.

4.2.3. Demonstrator for practical testingFor a future demonstrator there are two prin-cipally different track systems employing (a)discretely supported, and (b) continuously-elastically supported and embedded rails asshown in Fig. 3. Due to the low number ofconstruction elements and its quasi monolith-ic structure, the latter system is particularlywell suited for testing purposes.Practical testing of the early detection andtrack state monitoring equipment is per-formed using a modern Embedded Rail System

(ERS) in collaboration with Edilon Ltd. In anERS the rails are embedded in a two longitu-dinal troughs encarved in a concrete slab andelastically fastened by a two-componentpolyurethane-cork granulate-glue filling, calledcorkelast (see Fig. 3b). In contrast to thestandard ballast track (Fig. 1a) and most slabtrack systems derived from it, where the railsare only supported at discrete support points,in an ERS the rails are continuously-elasticallysupported over their entire length, both in ver-tical and horizontal directions (HOHNECKER,2002). Furthermore, instead of about 55,000parts (bolts, nuts, clamps) per km of track, inthe case of a ballast track, an ERS employsonly few structural components (see Fig. 3b)that are quasi monolithically connected.PVC pipes are usually inserted in order toreduce the required corkelast mass. If the sen-sors (e.g. strain gauges, accellerometers) arenot embedded in the corkelast itself, the sen-sors and their wiring can be accomodatedinside the PVC pipes. In this way they are pro-tected from vandalism and damage due totrain operation. Standardized units of ERS testtrack sections can be integrated in the world-wide railway network.

4.3. Advanced IaC technologiesFor the communication of seismological basisstations with central processing stations sever-al transmission methods (via satelite, spreadspectrum, ADSL) with high stability and effi-ciency in noise suppression can be employed.

Figure 3: (a) Standard ballast track with discrete rail support(b) details of a continuously- elastically supported Embedded Rail System (ERS)

(a) (b)

38

The choice of the method depends on the cir-cumstances (possible perturbations, distancebetween stations, available energy) of the par-ticular application. In the railway sector, thepossibility of using the standard GSM-R has tobe investigated.Usually, the communication between an earlydetection system and a traffic control centerwill involve different data transmission formatsdepending on the country where the earlywarning system is to be installed. In order toavoid frequent interface adjustments, theimplementation of a »message broker«, whichaccomplishes the required universal transfor-mations between different data formats usedby various standard and non-standard applica-tions is recommended.Spatial data plays an important role for theearly warning system for transportation lines.On the one hand spatial earthquake data(shake maps) need to be recorded, edited,stored, modelled and analyzed, as well asalphanumerically and graphically visualized.On the other hand this data needs to becombined with the spatial data of the trans-portation lines themselves. The early warningsystem shall take this tasks into account andshall make usage of available geospatialstandards, for example provided by the OGC(Open Geospatial Consortium) and shallmake usage of adequate functions of geo-graphical information systems (GIS) for therealization of the early warning system,which intends to minimize the risks causedby earthquakes.In this context, an important role is played bymodels that allow to simulate the behavior ofvarious infrastructure components of a trans-port line. Such models can be of differentnature, depending on the state of the art inthe particular field, i.e., they are either encod-ed heuristically as a set of rules or in mathe-matical form. The complexity of such modelscan be very high. For most infrastructure com-ponents one can use existing models (e.g. thevibrational behavior of a bridge, tunnels, orearthworks). New models have to be devel-oped for special tasks, e.g., a derailmentmodel. In any case, it is important to establish

an interoperability of various models and sim-ulation tools by implementing the standardHLA (High Level Architecture).For an integrated system component of a deci-sion support system, experts from the respec-tive transportation system must define therules and criteria, which if satisfied lead to aunique set of measures (early warning of acertain region, stopping of vehicles, blockingof certain road/track sections). This processleads to an increased transparency and justifi-cation of certain decisions. Furthermore, thedecisions become rational and reproducible. Ifthese prerequisites are satisfied, a computer-based decision support system can be con-structed through formalization of these rulesand criteria. Such a computer-based decisionsupport system is an essential part of EWSTransport, because one cannot expect that suf-ficiently fast and effective decisions can bemade by the responsible emergency managersgiven the boundary conditions listed in section4.2 they have to cope with. Also in the caseof an implemented IT solution the factor timeis decisive.Therefore an Information System for the EWSTransport System will be developed, whichtakes into account the tasks and aspects dis-cussed in the previous paragraphs. The archi-tecture of the Information System will bedesigned following the »Reference Model ofOpen Distributed Processing (RM-ODP)«. Thedesign foresees to look at the architecturefrom different viewpoints. The first viewpointdefines use case scenarios for the usage ofthe Information System, which shall be real-ized. The first version of the architecture ofthe Information System for the EWS TransportSystem distinguishes between the followinguse cases:1. Early warning system for transport lines2. Dynamic early warning system for transport

lines, which describes the workflow fromthe recognition of an earthquake to thefinal action, which needs to be performed(e.g. stop of a train)

3. Generation of an event report, which canbe created for the post-processing ofan event.

39

4. Generation of an damage report, which canbe created for the post-processing of anevent.

5. Training of the Artificial Neural Network,which is used to detect an earthquake.

6. Preparing of the damage catalogue, whichis needed for the evaluation of predicteddamages.

7. Continuous monitoring of infrastructureelements of transport lines

In the second viewpoint information modelswill be created. They will be used in the ser-vice specifications of the third viewpoint. Onegoal will be to reuse and adapt alreadydefined information models and service speci-fications from consortiums or projects like theOGC or ORCHESTRA (European IntegratedProject, which has developed a referencearchitecture for the application field »RiskManagement«). The models of viewpoints twoand three are described in abstract UML dia-grams before the mapping to current tech-nologies is implemented. This ensures that thearchitecture can be flexibly adapted to newlydeveloped technologies in future.

5. ReferencesAllen, R. and Kanamori, H., 2003. The poten-tial for earthquake early warning in South Cal-ifornia, Science, 300, 786–789.

Berger, R. et al., 2005. The way to coordinat-ed deployment of ERTMS/ETCS throughoutthe European Network, Railway TechnicalReview 4, pp 21.

Böse, M., Erdik, M. and Wenzel, F., 2005.Earthquake Early Warning – Real-time predic-tion of ground motion from the first secondsof seismic recordings. Proc. Volume of the Int.Conference on 250th Anniversary of the 1755Lisbon Earthquake, 185–187, 1–4 Nov. 2005;Near real-time estimation of ground shakingfor earthquake early warning. In: Malzahn, D.& Plapp, T. (eds). Disaster and Society – FromHazard Assessment to Risk Reduction, LogosVerlag, 175–182, 2004.

Erdik, M., Fahjan Y., Ozel O., Alcik H., Mert A.,and Gul M., 2003. Istanbul Earthquake RapidResponse and the Early Warning System. Bull.Of Earthquake Engineering, V.1, Issue 1, pp.157–163.

Espinosa-Aranda, J., Jiménez, A., Ibarrola, G.,Alcantar, F., Aguilar, A., Inostroza, M., andMaldonado, S., 1995. Mexico City seismicalert system, Seism. Res. Lett. 66, 42–53.

Hohnecker, E., 2002. Diskret gelagerte oderkontinuierlich eingebettete Schienenfahrbahn-systeme? EI-Eisenbahningenieur 53, 45.

Hohnecker, E. 2003. Acoustic properties ofrailway superstructures, World Congress onRailway Reserach, 28 Sept.–1 Oct. 2003, Edin-burgh.

Nakamura, Y., 2004. UrEDAS, Urgent Earth-quake Detection and Alarm System, Now andFuture, 13th World Conference on EarthquakeEngineering, Vancouver, August 1–6, 2004.

Quante, F., 2001. »Innovative Track Systems –Criteria for their Selection«. Better Railwaysfor a European Transport Market, Quante, F.(Ed.) Media Network, Pfungstadt, Nov. 2001.

Quante, F., Kuhla, E., Strothmann, W., 2005.»CroBIT – Cross Border Information Technolo-gy for Interoperability in European Rail FreightTransport«. Proceedings 5th European Con-gress on Intelligent Transport Systems, Han-nover, Juni 2005.

Röthlingshöfer, F., 2002. Erschütterungstech-nische Analyse eines Straßenbahnoberbaus,Fallstudie Thereses gate, Oslo, Diplomarbeit,Universität Karlsruhe.

Usländer, T. (Ed.), 2005. »Reference Model forthe ORCHESTRA Architecture (RM-OA)«. OpenGeospatial Consortium Discussion Paper(05–107), 2005; »Trends of environmentalinformation systems in the context of theEuropean Water Framework Directive«. ELSE-

40

VIER Journal Environmental Modelling & Soft-ware 20 (2005) 1532–1542.

Teng, T. L., Wu, Y. M., Shin, T. C., Tsai, Y. B.,and Lee, W. H. K., 1997. One minute after:strong-motion map, effective epicenter, andeffective magnitude, Bull. Seism. Soc. Am. 87,1209–1219.

Tsai, Y. B. and Y. M. Wu, 1997. Quick deter-mination of magnitude and intensity for seis-mic early warning, 29th IASPEI meeting, Thes-saloniki, Greece.

Wu, Y. M. Shin, T. C., and Tsai, Y. B., 1998.Quick and reliable determination of magnitudefor seismic early warning, Bull. Seism. Soc.Am. 88, 1254–1259; Wu, Y. M., Chung, J. K.,Shin, T. C., Hsiao, N. C., Tsai, Y. B., Lee, W. H.K., and Teng, T. L., 1999. Development of anintegrated seismic early warning system in Tai-wan – case for the Hualien area earthquakes,TAO 10, 719–736.

Wenzel, F., Oncescu, M.M., M. Baur, F. Fiedrich& C. Ionescu, 1999. An early warning systemfor Bucharest, Seismol. Res. Letters, 70, 2,161–169.

Wenzel, F., Baur, M., Fiedrich, F., Oncescu,M.C. & Ionescu, C., 2001.Potential of Earth-quake Early Warning Systems, Journal of Nat-ural Hazards, 23, 407–416, 2001.

Wenzel, F. and Marmureanu, G., 2006. Earth-quake Information Systems, to be published inPageoph Topical Issue, Proc. Volume 22nd Int.Tsumani Symposium, Chania, Greece, 27–29June, 2005.

41

METRIK – Model-Based Development ofTechnologies for Self-Organizing DecentralizedInformation-Systems in Disaster Management(DFG-Graduiertenkolleg)

SummaryRecent progress in basic research has lead tovisions how to use new self-organizing net-works for advanced information systems.These networks function without central ad-ministration – all nodes are able to adaptthemselves to new environments autonomous-ly and independently. The addition of newnodes or failure of individual nodes does notsignificantly impact the network’s ability tofunction properly. Information systems andunderlying technologies for self-organizingnetworks, in the context of a specific applica-tion domain, are the central topic of researchfor this graduate school.The research focuses on the important tech-nologies needed at each individual node of aself-organizing network. Research challengeswithin this graduate school include: finding apath through a network with the help of newrouting protocols and forwarding techniques,replication of decentralized data, automateddeployment and update of software compo-nents at runtime as well as work-load balanc-ing among terminal devices with limitedresources. Furthermore, non-functional aspectssuch as reliability, latency and robustness willbe studied.The graduate school’s focus on decentralizedinformation systems with self-organizing net-

works is sharpened by relating those moregeneral research issues to a very specific appli-cation domain: computer-supported disastermanagement. For this reason, the graduateschool emphasizes the use of techniques,methods, and concepts for designing andimplementing geographic information systemson top of dynamic, highly flexible, self-orga-nizing networks and their integration with ser-vices for geographic information systemsbased on existing technologies in these areas.To manage the complexity of data, informa-tion, and services and to make them availablefor the user, it is extremely important to hide(as much as possible) the difficulties and thecomplexity of such an environment. Only if wesucceed in shielding the user from internalerrors and/or changes, such systems will beaccepted.Aside from the specific demands given by thegeographically dispersed positions in our appli-cation domain, the network topology plays animportant role in the configuration process,because the partitioning into separated sub-networks must be detected and prevented.Furthermore, if the network should have beenpartitioned, it must be possible to find all sub-networks and reconnect them – if possible –with a minimal number of links.

Speaker: Fischer J. (1), [email protected]

(1) Institute for Informatics, Humboldt University of Berlin

(2) GeoForschungsZentrum Potsdam

(3) Zuse Institute Berlin

(4) Fraunhofer FIRST

(5) Hasso Plattner Institute for Software Systems Engineering

42

Research in the suggested application domainis interdisciplinary by nature. The graduateschool will conduct basic research in applyingworkflow management technology to disas-trous events, such as earth quakes, based onthe developed network protocols and infor-mation service concepts. The goal of this workto support decision makers in making betterinformed decisions by using the completerange of available options.A key differentiator of this graduate school isits model-based approach that will be appliedto all layers of the system. Meta-model lan-guages will aid disaster management expertsto model their workflows, which may in turnbe simulated in order to assess decisionprocesses. Theoretical studies of workflowusability will provide the basis for investiga-tions of the composability of partial work-flows in complex scenarios. Workflows willalso be studied for their applicability to aidthe self-organization of systems by dynami-cally allocating network resources. The com-bination of functional specification, automat-ed code generation, and performance analy-sis methods is a distinguishing aspect ofmodel based service engineering in self-orga-nizing distributed information systems, whichwill contribute significantly to the servicequality of all system components.

Current PhD Subjects

Transactional Workflows in Self-OrganizingInformation SystemsArtin Avanes, [email protected]

The integration of sensor networks, otherembedded systems and mobile devices build-ing a self-organizing information system (IS)results in a faster detection, a better preven-tion and recovery from urban disasters. How-ever, an efficient disaster management strong-ly depends on the reliable execution of work-flow processes before, during and after a dis-aster strikes. Workflows describe the opera-tional aspects of work procedures specifyingthe individual tasks, the order, and the condi-tions in which the tasks must be executed. In

the area of disaster management, workflowactivities may run on sensor networks measur-ing the ground motions and providing mobilerescue teams with important event data. Theintegration of smart, wireless collaboratingdevices poses new technical challenges for thecoordination and execution of workflow activ-ities. Unlike traditional IS, we have to copewith– Dynamics: Device nodes may fail or the com-

munication links between devices can tem-porally break.

– Limited Resource Capabilities: Self-organiz-ing networks contain smaller devices, withlimited resources, e.g. limited battery poweror a finite memory capacity.

– Heterogeneity: Different kinds of partici-pants with varying software and hardwarecapabilities have to be considered during theworkflow execution.

In view of these challenges, we will focus onstrategies to support an energy-efficient distri-bution and a robust execution of workflowactivities in such self-organizing environments.Thereby, we use the approach of constraintprogramming to efficiently distribute the activ-ities. We model the deployment task (mappingof activities to device nodes) as a constraintsatisfaction optimization problem (CSOP). Ourgoal is to select the optimal assignment con-cerning the energy consumption from the setof possible execution settings. Based on atransactional model for workflows relaxing theACID properties, we will also develop differentfailure protocols, such as replication andmigration strategies. We will compare thecosts for migration of activities with the repli-cation efforts for different failure classes (soft-ware, hardware, communication failures).

Dependable, Service-Based Infrastructurefor Disaster ManagementStefan Brüning, [email protected]

Computer systems for use in disaster manage-ment are based on very heterogeneous infra-structures. Powerful servers are the backboneof those systems which are used for workflowmanagement, task force coordination and

43

communication. These servers process datagenerated by humans and sensors. Sensordata can come from a variety of different sen-sor types mounted to individual sensor nodes.The complexity of these sensors ranges fromsophisticated physical systems to simple nodesas used in sensor networks that have only verylimited resources.All parts of the system have to work effectivelyand efficiently. A special focus lies on depend-ability. A break down of some parts must notlead to a failure of the whole system butinstead must be compensated by the remain-ing parts. The system has to self-adapt to thechanged infrastructure and still allow success-ful management of the disaster.Service-oriented architectures (SOAs) havebeen established as a standard for distributedsystems in recent years. The possibility todynamically discover and bind to new servicesallows a loosely coupled and adaptive infra-structure. The dependable integration of pow-erful servers and resource limited sensor nodesinto one disaster management system is achallenging task.Servers for use in workflow systems can beembedded in a service- based architecturebased on web services. Several approaches forincreasing dependability for web services exist.Nevertheless, there are no exact measure-ments yet whether those increases in depend-ability can really be reached.In my dissertation I will examine to whichextend certain technologies and methods areincreasing dependability. For this purpose, aSOA will be deployed in our laboratory whichwill consist of several test services on differentcomputers. Artificial failures will be simulatedusing fault injection technologies. The resultswill be evaluated based on metrics for servicedependability. Additionally a fault taxonomywill be developed as a basis for fault injection.

Modeling and Verifying DeclarativeWorkflowsDirk Fahland, [email protected]

The support of disaster management by self-organizing information systems and networks

cannot succeed without support of the inter-meshed work procedures emerging in thiscontext: Administrative processes that are car-ried out when disaster strikes and that coverthe overall situation need to fit to emergencyand rescue procedures for locally limitedactions. Both will interfere with proceduresand behaviour of the supporting self-organiz-ing information system. A failure free androbust concurrence, that is correctness of theinvolved procedures, is prerequisite for suc-cessful disaster management.Workflows have been established as a mathe-matically founded, operational model for workprocedures in different variants. Thus ques-tions on the correctness of one or several jointprocesses can be formulated and answered ona principle basis. As any model, establishedworkflow models have been built underassumptions like availability of resource, com-munication infrastructure and continuity ofcooperating partners. The dynamics of disastermanagement and self-organizing systems vio-lates these assumptions: Workflows for disas-ter management need to be flexible to allowthe execution of a process in various circum-stances. Unpredictable events require the needfor (runtime) adoption of a workflow. At thesame time, a flexible and adaptive workflowhas to be correct in order to be a reliable toolin disaster management. A property of inter-est will be the self-stabilization of a workflowthat encountered a faulty situation.We identified two ideas that may effectivelyallow the modeling of flexible and adaptiveworkflows such that these models can be ana-lyzed for properties of interest. We step backfrom classical operational models like Petrinets and introduce declarative and scenario-based elements: Temporal logics, for instancelinear-time temporal logic, allow to preciselymodel critical aspects like communication orresource access while giving only a loose, butsound, characterization of ordering tasks ofworkflow. A scenario-based approach as exer-cised in Harel’s Life-Sequence Charts breaksbehaviourally complex models into human-conceivable parts that can be modified as the(changing) situation requires.

44

This approach requires the definition of a fea-sible (declarative and/or scenario-based) mod-eling language for workflows that meets theideas and concepts of processes in disastermanagement. Having models in such a lan-guage, the models have to be made effective,for instance by translating them into an oper-ational model (and vice versa). Finally, we areinterested in a (temporal-) logic-based charac-terization of relevant properties like self-stabi-lization and feasible verification procedures toverify the correctness of a workflow.

Data Management for Wireless SensorNetworksTimo Mika Gläßer, [email protected]

The Graduierten Kolleg METRIK focuses onself-organizing systems for use in disastermanagement before, during and after nat-ural disasters like earthquakes, tsunamis orvolcanic eruptions. We will try to addressthe various problems that arise with a) datagathering and b) event-detection usingwireless sensor networks in the above men-tioned scenarios.In fact wireless sensor networks play a centralrole in the envisioned systems for disastermanagement. Wireless sensor networks con-sist of small, independent sensor devices withhighly incomplete and local knowledge thatcan measure physical properties of their envi-ronment and communicate with other sensordevices in their area using broadcast wirelesscommunication. These sensor devices areequipped with small batteries as their onlypower-source. However each sampling opera-tion, each computation and communicationoperation on the sensor devices consumesenergy. Hence respecting the energy con-straints of these wireless sensor devices is crit-ical to keep them operational and optimizingfor lower energy consumption may increasedevice and network lifetime.We will try to provide a declarative publish/subscribe-style interface to the sensor networkthat enables users to specify queries like»Report all locations of sensors in the areaaround Istanbul where the shaking exceeds a

threshold τ.« or »Report all clusters of nodesand the average shaking in the clusters in theIstanbul area where at least 80an earthquakeof magnitude greater than τ.«When a user hands such a query to our sys-tem, the query will automatically be decom-posed into smaller operations arranged in adataflow graph and be spread across the wire-less sensor network.The operations in the dataflow graph can beexecuted in different orders yielding differentenergy usages at both the individual sensordevices and the network level. Our system willcontinuously try to optimize the cost of multi-ple such queries to extend the wireless sensornetwork’s lifetime. Additionally as routing is ofutmost importance in multi-hop wireless net-works we provide novel ideas of includingrouting decisions in the evaluation and opti-mization process of queries.Prior research in the area does either not per-form in network query processing – all data isrouted to a gateway node and processed there– or considers single-queries only. Additionallyour approach differ from others by not onlyallowing simple aggregations like MIN, MAX,AVG but also supporting the clustering ofnodes based on the network topology, e.g.hop-distance.

QoS in 802.11 Multihop NetworksKai Köhne, [email protected]

Wireless local area networks using the IEEE802.11 standard are popular both in homeand office use. Most often, the technology isused for the »last hop«, letting the useraccess the (wired) core network via an accesspoint. However, projects such as MIT Roofnetand Berlin Roofnet demonstrate that even alarger metropolitan network can rely solelyon wireless links. In these networks, everyparticipating node does not only process itsown data, but also relays traffic for neigh-bouring nodes, so that every node in the net-work can be reached by multi-hop connec-tions. These techniques have also been stud-ied and examined for sensor networks. In theEDIM and SAFER projects we create an earth-

45

quake early warning system based on 802.11multi-hop technologies.Implementing demanding real-time applica-tions such as earthquake early warning basedon wireless multi-hop networks require highnetwork reliability. However, each link in a802.11 network is highly volatile, sufferingfrom physical phenomena like fading, but alsofrom interferences with other devices. More-over, in cities often 10 or more 802.11 net-works operate in parallel, effectively limitingthe available bandwidth for each of them.One way to achieve a higher reliability for timecritical applications is service differentiation.The recent addition 802.11e provides newmedium access methods that share the medi-um either in a fully distributed (EDCA) or cen-tralized (HCCA) way. EDCA is of special impor-tance for multi-hop networks: High prioritypackets have a higher chance to access themedium in the case of conflicts.In my work I evaluate the advantages andproblems of using EDCA in wireless multi-hopnetworks. One problem largely ignored in theliterature is spectrum competition betweendifferent 802.11 networks. Networks are ableto optimize their EDCA parameters so thatthey can gain a higher share of the band-width. If every network does use the mostaggressive settings, the differentiation mecha-nism is effectively undermined. We thereforedevelop a new differentiation protocol whichbuilds upon EDCA. In these protocols, everydata message has the same EDCA priority. Ser-vice differentiation is achieved using a sepa-rate control channel. This avoids the manglingof intra- and inter-network priorities, andresolves unfairness issues due to interferencebetween data and control packets.

Semantic Integration for DisasterManagementBastian Quilitz, [email protected]

Fast and effective disaster management large-ly depends on the fast provision of detailed,precise, and up-to-date information. It is vitalfor decision makers and rescue crews to beinformed about the state of critical infrastruc-

tures, the number, location, and status ofinjured people, etc. The required informationis likely to be scattered over many distributedinformation systems, belonging to differentorganizations and not all relevant informationsources may be known in advance. Anotherimportant aspect in disaster management isthe different needs depending on the user’srole. Decision makers will require differentinformation than rescue crews in the field.Integrated information systems provide accessto a number of information sources. In thosesystems heterogeneity is one of the main chal-lenges. While there exits some approaches toovercome technical heterogeneity, such asobject oriented middleware, overcoming struc-tural and semantic heterogeneity is still anongoing research question.For disaster management we distinguishbetween two types of information sourcesbased on their temporal characteristics: Data-bases and streaming information sources.While databases provide query access to per-sistent data that is relatively static, such asbackground information or management in-formation, streaming information sources areupdated continuously. Examples for streaminginformation sources are sources for sensordata or locations of rescue teams.In the context of METRIK, I envision an infor-mation system for disaster management thatseamlessly integrates both types of informa-tion sources and provides the right informa-tion to the right user. Users can either submitsubscriptions to the system and will continu-ously receive updates or ask ad-hoc queries ifthey have special information need. In thiscontext, I will investigate methods for seman-tic integration of geo data from a multitude ofheterogeneous, distributed data sources. Thefocus will be on efficient query processing forthe integration system, utilizing the spa-tiotemporal context of the data for query opti-mization. The goal is a timely provision of inte-grated, spatiotemporal consistent data. Sup-plementary, I will investigate methods toensure data quality.

46

Using metaprogramming and metamodelingfor prototyping middleware abstractionsin wireless sensor networksDaniel A. Sadilek, [email protected]

In METRIK, wireless, self-organizing sensornetworks (WSNs) are supposed to be the basisof complex disaster management systems.Software development for such systems is dif-ficult because of numerous reasons: applica-tions on WSNs are distributed and show theusual synchronization and communicationproblems; they have to respect the limitedresources of the sensor nodes; deploymentand debugging is very costly. Middlewarecould help with these problems but existingmiddleware technologies are inappropriate forWSNs – and it is not yet clear which services,which programming model and which abstrac-tion level a middleware technology for WSNsshould offer. Furthermore, for WSNs the term»middleware« is used in a much broadermeaning than traditionally: there are operatingsystems, languages and pure libraries allnamed »middleware«.I am looking at middleware from the per-spective of domain specific languages (DSLs),which may enable domain experts to activelydevelop parts of a WSN application. Becauseservices, programming model and abstractionlevel are not yet clear, I am working on tech-nologies that allow the prototyping and sim-ulation of a DSL’s domain concepts and itssemantics. The definition and usage of DSLs isroutine since decades in metaprogrammablelanguages like Lisp and Smalltalk. However,these languages lack support for purpose-built concrete syntax. On the other hand,technologies from the field of model-drivensoftware engineering (MDSE) provide suchsupport. However, they miss a thing thatmetaprogrammable languages provide: directexecutability – because metamodels alonedon’t have operational semantics. The seman-tics is usually provided by a transformation, inwhich the knowledge how to map domainspecific concepts to concepts of the targetplatform is encoded.

My approach is to use flexible metapro-grammed domain abstractions in which fixedobject-oriented metamodeling layers are intro-duced. I hope to gain synergies from this com-bination that allow a great part of the lan-guage semantics done for the simulation to bereused for the target platform. Currently, I amdoing a proof of concept implementationbased on the Eclipse modeling technologiesand Scheme. I am working on a developmentprocess specification, the simulation of multi-ple nodes, support for communication primi-tives, configuration of runtime parameters,support for mixed graphical/textual DSLs, andcompilation for the target platform. As firstevaluation project, I am developing a stream-based language, which can be used for thedevelopment of an earthquake early warningsystem with algorithms for earthquake detec-tion, distributed warning determination, andwarning dissemination.

Language ModelingMarkus Scheidgen, [email protected]

Computer languages are in the centre of soft-ware engineering; they are the tools to com-municate problems and solutions to bothhumans and computers. Computer languageshave to allow unambiguous and effectiveexpressions, independent of the specific lan-guage nature, its level of abstraction, purpose,or area of application. Language sentenceshave to be proofed and tested. The artefactswritten in computer languages have meaning;these semantics is manifested in compilers,simulators, model transformations, or employ-ing calculi and formalisms. In this sense, acomputer language is much more than a setof sentences; it is more a set of tools that pro-vide for all these aspects.Language modeling is based on the hypothe-sis that languages are pieces of software andthat they should be developed like software ina model driven fashion. We use models todescribe languages, their concepts, notations,and semantics. Language models, known asmeta-models, are artefacts written in severalmeta-languages. Each of these languages can

47

be used to describe a single language aspect.These aspects include, for example, abstractlanguage structure, static semantics rules, tex-tual or graphical notations, operationalsemantics or code generation. The semanticsof a meta-language is manifested in a generictool which uses a language description writtenin that meta-language. The ultimate goal is toprovide a complete framework of metalan-guages that allow to describe a language in allnecessary aspects. Whereby, these descriptionscan be used by humans to understand the lan-guage and by machines to provide automatedtool support for the language.We contribute several meta-languages andgeneric tool support for the modeling and useof computer languages. We developed themeta-modeling framework A MOF 2 for Javabased on the MOF 2.0 standard. This frame-work, in contrast to others, supports en-hanced refinement and specialisation featuresthat allow for more flexible and reusable struc-ture models. It also uses a new Java mappingwhich utilises generics and variant return typesfor safer programming with models. We devel-oped the Textual Editing Framework, whichprovides a meta-language for textual notationmodeling. The framework includes a genericeditor that provides semantic rich editing withsyntax highlighting, code completion, errorannotations, and much more. A third contri-bution provides meta-language and generictool support for execution semantics. Basedon structure definitions for abstract languagesyntax and runtime data, the language engi-neer can specify behaviour at the meta-level.The generic model simulator can then executelanguage instances solely based on suchdescriptions. We are evaluating on the Speci-fication and Description Language (SDL). Thisproofs the general applicability (and also theweaknesses) of our contributions and is ourbasis to reason about the hypothesis: lan-guages, as software, can be modeled.

Model Driven Engineering (MDE) for Modelingand Simulation in Disaster ManagementFalko Theisselmann,[email protected]

Scientific models and simulations of environ-mental systems are major information sourcesin disaster management. Various frameworksand domain specific languages have beendeveloped to ease modeling and simulationtasks and provides access to new modelingand simulation technologies to non-experts.These frameworks usually stick to certainmodeling formalisms (i.e. box models, statemachines, cellular automata etc.) and model-ing languages. As a result, models are usuallyframework and language specific. Usingframeworks requires knowledge about frame-works, their concepts, languages, and capabil-ities. Changing the framework usually requiresexpensive, manual reimplementation of mod-els. In environmental modeling practice, this isan obstacle for the use of frameworks, thereuse of models, and their integration in cou-pled multi-model models.We investigate the use of MDE to tackle thisproblem. Our approach is conceptually basedon the Object Management Group’s ModelDriven Architecture (MDA). Usual modelingformalisms in the disaster managementdomain are platforms in the sense of MDA.Domain specific models (DSMs) are mappedonto formalism specific models (FMSMs). Inthe following steps, these are mapped ontoframework specific models (FWSM) that allowfor the generation of framework specific exe-cutable code.One DSM maybe transformed into severalFMSM. Note that one FMSM maybe trans-formed into other FMSMs and into severalFWSMs. This yields the possibility for theextensive reuse of models, if the formalismsallow for that. Also, several DSMs maybetranslated into a single FMSM, which facili-tates model integration/coupling.To formulate DSMs, we use DSLs, where theconcepts and notational elements are tai-lored to the concepts and cognitive spacesof the domain experts. The logic defined

48

by the transformations ensures the correctapplication of formalisms and conceptswhich otherwise would be the responsibilityof the domain expert.The goal of our research is to evaluate andshow how MDE may enhance model reuse,integration and combination of differentdomain models and promote the sensibleuse of frameworks. Moreover, we wantto investigate how to represent domainknowledge adequately and support collabo-rative, multidisciplinary modelling in the fieldof disaster management.

Metamodel-based Language EngineeringGuido Wachsmuth, [email protected]

In METRIK, we follow a model-based approachto develop disaster management systems. Wewant to provide domain experts with model-ling means that are intuitive, concise, andsemantically precise. To fulfil these require-ments, we propose a combination of severaldomainspecific languages (DSLs). LanguageEngineering brings Software Engineering tolanguages. It is concerned with language de-sign, language maintenance, language exten-sion, language recovery, translation, genera-tion, interpretation, etc. Metamodels are ourchoosen tool to specify languages. Like othersoftware artefacts, metamodels and meta-model-based language descriptions evolveover time. Metamodel evolution is usually per-formed manually by stepwise adaptation. Fur-thermore, models need to co-evolve in orderto remain compliant with the metamodel.Without co-evolution, these artefacts becomeinvalid. Like metamodel evolution, co-evolu-tion is typically performed manually. This error-prone task leads to inconsistencies betweenthe metamodel and related artefacts.In our work, we explore transformationalmetamodel adaptation as a foundation forMetamodelbased Language Engineering. As afirst result, we provide a theoretical basis tostudy the effects of metamodel evolution interms of metamodel relations. We employwell-defined evolutionary steps for metamod-els compliant to OMG’s Meta Object Facility

(MOF). The steps are specified as transforma-tions in QVT Relations, the relational part ofOMG’s Query-View-Transformation language.Each step forms a metamodel adaptation andis classified according to its semantics andinstance-preservation properties. Automaticco-evolution steps are deduced from thesewelldefined evolution steps. This co-adapta-tion prevents inconsistencies and metamodelerosion. Starting from our theoretical results,we develop an Adaptation Browser for MOFcompliant metamodels. The tool is built uponthe Eclipse Language Toolkit. This provides uswith undo/redo support, adaptation history,and scripting facilities.We now examine our approach in two casestudies. The first study is concerned with thedesign of a new DSL. The transformationalapproach facilitates a well-defined stepwisemetamodel design. Starting from basic fea-tures, new features are introduced by con-struction. We hope that extensive usage ofthis principle leads to an agile process. Thesecond study is concerned with the extractionof a language out of a natural language doc-ument. The document describes disaster man-agement processes semi-formally. Languagerecovery is concerned with the derivation of aformal language specification from suchsources. For grammar recovery, a transforma-tional approach already proved to be valuable.In a similar way, we employ our approach toassist metamodel recovery.

Model Coupling and TestStephan Weißleder,[email protected]

Models are a fundamental element of model-driven software development (MDSD). Byusing model transformations and code gener-ation patterns, available tools are already ableto generate a large part of the source codeautomatically. However, while effort has beeninvested in source code generation, the auto-matic generation of test code has been wide-ly neglected. This test code should revealerrors, which are occurences from abnormalprogram terminations to small deviations from

49

the specified behavior. Normally, the latedetection of such errors results in high costsfor their removal and for the compensation oftheir consequences. So, early and extensivetesting is most important. Nevertheless, thepriority of testing is often still considered lowerthan the priority of delivering products early,which is mainly due to high costs for test suitedevelopment.All in all, models and tests both aim at improv-ing the quality of the developed system. Ourcurrent work combines methods from thesetwo fields.The intention of this combination is to gainextra benefit for software testing by using dif-ferent models together. Therefore, we use sev-eral models describing different aspects of asystem, combine them, and derive enoughinformation for automatic test case genera-tion. Several former approaches only dealtwith single models or with models that usedjust a very small aspect of another model.Likewise, constructive but random approachesfor automatic test input data generation havebeen neglected.Our current state of affairs includes the auto-matic and model-based generation of a wholetest suite based on state machines and classdiagrams of UML. We combine both modelsvia their OCL constraints and construct parti-tions for the corresponding test input values.Furthermore, we operate boundary testing onthese partitions to generate test inputs forpositive and negative test cases. Additionally,our current work combines structural andbehavioral models to benefit from their partic-ular relationships. For instance, behavioralspecifications (e.g. state machines) can bereused in several static contexts (e.g. classes)via structural relationships like inheritance. Wealready implemented our main ideas inSMOTEG – a tool based on the Eclipse model-ling framework.The identified challenges of our currentapproach are the following: the identificationof retraceable constraints so that the influenceof input parameters becomes clear; the com-bination of several behavioral specifications forone class (multiple inheritance); and the intro-

duction of concurrency, which is important fordistributed systems like sensor nets. In thecontext of METRIK, we aim at adapting ourapproach to special tasks of geo informationsystems (GIS) or meta-modeling. Furthermore,we need to identify significant case studies toshow our approach’s benefit.

Associated PhD Subjects

Model-based simulation of decentralized,ad-hoc sensor systemsFrank Kühnlenz,[email protected]

The popularity of sensor systems increases withthe ability of realizing more complex scenariosat low costs. Additionally the hardwarebecame cheaper and even more powerful overthe last years. This leads to the vision of »smartdust« whereby billions of smallest sensors pro-vide services in a cooperative way. They forman ad-hoc network where sensor nodes mayfail and new ones may be added and act coop-eratively in a self-organizing and »intelligent«way (e.g. distributed computing).One challenging application domain for sensornetworks are earthquake early warning sys-tems (EWS). Especially the reliability of the sen-sor net is a crucial point in this domain whereunder disadvantageous conditions single sen-sors may be destroyed but the whole systemcan detect the earthquake nevertheless. Todaysensor nets for earthquake early warning havea static infrastructure which was carefullyplanned by an operator and comprises one ormore data centers (e.g. Turkey IERREWS). Inthis case the data center does all computingand decision logic.In two international, interdisciplinary projects,SAFER and EDIM, methods and technologiesare developed and prototypically realized foran earthquake early warning system basedon a decentralized, wireless, mesh sensornetwork. Like every early warning system itmust match several optimization goals: e.g.early warning time as the most importantone but also minimal false positive alarms(which should be guaranteed by issuing an

50

alarm only after »discussion« within a groupof sensor stations).For this recent approach new distributeddetection algorithms and consensus basedalarm handling protocols must be created andexplored. Therefore my work concentrates ondeveloping model-based simulations and eval-uating them against specific criteria for EWS.We follow a top-down approach starting withhigher order questions – in case of alarm han-dling for example how is a group of sensorstations defined and how many stations arenecessary. According to that we first abstractfrom the transport oriented network level andinstead find constraints for the network layer.Such a constraint could be the conclusion: If anumber of messages between designated sta-tions can be transported within a definedtime, then the early warning time is below aspecific value.To follow the top-down approach and for sup-porting the abstractions of single levels, differ-ent simulation frameworks are used and mustbe coupled. For example models of communi-cation systems can be designed in SDL-RT verywell and directly executed (therefore it is usedfor modeling alarm handling). In the lastdevelopment steps it is necessary to couple aspecialized network simulation framework(e.g. ns-2).

Quality assessment of reports from the publicfor flood disaster managementKathrin Poser, [email protected]

For disaster management it is important tointegrate information from different sources toget an assessment of the situation which is ascomplete as possible. Particularly when largerareas are affected, the local population cancontribute valuable information for disasterresponse and recovery. New internet technolo-gies facilitate fast and easy collection of datafrom the public. So far, however, informationprovided by the affected population has onlybeen taken into account systematically inearthquake intensity mapping, and no com-prehensive evaluation of the quality of thesedata has been performed.

In order to make informant reports useful fordisaster management and use them as inputfor the estimation of damage, the followingissues need to be addressed:– Which types of information, that are requi-

red in disaster management, can be suppliedby the affected population?

– How can these data be collected?– How can these data be validated and their

quality be assured?– How can these data be integrated with other

information such as sensor data or remotelysensed data?

The research questions will be addressedexemplarily for the case of (rapid) flood lossestimation by empirical loss models whichdescribe the susceptibility of elements at riskto flooding. Input data for these models areinformation about the flooding event andabout the exposed values. The output is theestimated monetary loss.The use of information supplied by non-experts for disaster management is impededby the fact that its reliability is in generalunknown. Therefore, it is important to assessthe reliability and quality of these data. Ourapproach for such a quality assessment willrely on geostatistical methods and concepts oftrust models. These methods will take intoaccount a comparison of the statements witheach other as well as with other data aboutthe flood event such as gauge measurements,remotely sensed data, and a plausibility checkbased on terrain and topographic data.The results of this research will contribute tothe efforts to tap the potential of »humans assensors«, i.e. to make use of information pro-vided by the general public via the internet fordisaster management.

51

Earthquake Disaster Information Systemfor the Marmara Region, Turkey (EDIM)

AbstractEDIM – Earthquake Disaster Information Sys-tem for the Marmara Region – is an interdis-ciplinary research project that focuses onimproving earthquake early warning capacityin the Marmara region around the mega cityof Istanbul. The consortium that addresses thisgoal consists of the Geophysical Institute ofthe University of Karlsruhe (TH), Geo-ForschungsZentrum Potsdam (GFZ), the Com-puter Science Department of Humboldt Uni-versity – Berlin (HU), lat/lon GmbH, Bonn, DEL-PHI InformationsMusterManagement GmbH,Potsdam, and Kandilli Observatory and Earth-quake Research Institute of Bogazici Universityin Istanbul (KOERI).

Objectives of EDIMThe existing earthquake early warning systemfor Istanbul (Istanbul Earthquake RapidResponse and Early Warning System – IER-REWS) will be expanded to a regional scale(the Marmara Sea Region) and be significant-ly improved in terms of the quality of theearly warning information. Estimates of thereliability of the warnings will be delivered, aswell as real-time information dealing with theearthquake source parameters and near-real-time shake maps for the region. At the sametime, this project will serve as a test sitewhere new advanced sensor technology con-

sisting of self-organizing networks will beassessed. Data and information exchange willbe controlled by a dynamic geo-informationinfrastructure utilizing a user-oriented infor-mation and visualization strategy. The expect-ed results and products that have the poten-tial of being transferable to other earth-quake-prone regions are: real-time seismolo-gy using neural nets, self-organizing sensortechnology, real-time information systemsand personalized visualization systems. Wecan take advantage of a large body ofprevious work on ground motion assessment,site effect quantification, and risk assess-ment (KOERILOSS).On this basis, we address real-time informa-tion before, during and after an earthquake,following an interdisciplinary approach. Seis-mology will provide novel algorithms for therapid detection of earthquakes, source proper-ties, and the expected level of ground shaking,seismologists, computer and communicationscientists will develop the new technology ofself-organizing networks for early warning andrapid reporting, while geoinformatics will pro-vide methods for the incorporation of addi-tional information and will focus on tuning theresulting early warning information to a user’sneeds.In order to achieve these objectives, we havebuilt a consortium (Verbundvorhaben) of

Wenzel F. (1), Erdik M. (2), Zschau J. (3), Milkereit C. (3), Redlich J. P. (4),

Lupp M. (5), Lessing R. (6), Schubert C. (6)

(1) Universität Karlsruhe (TH), Geophysikalisches Institut, Hertzstr. 16, 76187 Karlsruhe

(2) Kandilli Observatory and Earthquake Research Institute, Bogazici University, Cengelkoy,

Istanbul, Turkey

(3) GeoForschungsZentrum Potsdam (GFZ), Telegrafenberg, 14473 Potsdam

(4) Humboldt-Universität, Institut für Informatik, Unter den Linden 6, 10099 Berlin

(5) lat/lon GmbH, Aennchenstr. 19, 53177 Bonn

(6) DELPHI InformationsMusterManagement GmbH, Dennis-Gabor-Str. 2, 14469 Potsdam

52

research and commercial organizations con-sisting of– the Geophysical Institute, University of

Karlsruhe (TH),– GeoForschungsZentrum Potsdam (GFZ)– Computer Science Department, Humboldt

University of Berlin (HU)– lat/lon GmbH, Bonn– DELPHI InformationsMusterManagement

GmbH, Potsdam– Kandilli Observatory and Earthquake

Research Institute, Bogazici University,Istanbul, Turkey (KOERI)

The consortium is coordinated by TU, whileKOERI will allow access to the existing IER-REWS system, will implement our results andwill assist in the development of an improvedcommunications system. Data from the self-organizing sensors, combined with additionalspatial information, will be provided by aninteroperable information infrastructure. Theproject is structured into three packages:Real-time information from a regionalaccelerometer network (A), Self organizingsensor system (B1), Infrastructure for self

organizing sensor system (B2), InformationSystem (C).The specific objectives of the project are:(1) Expansion of real-time communication viasatellite within the existing system to theMarmara Region (Fig. 1). This step is crucial, asthe entire region is highly industrialized anddensely populated. The work will be financedby the Governate of Istanbul within 2007 and2008. From the very beginning, KOERI will pro-vide internet access to the IERREWS to itsGerman partners.(2) Optimization of the existing early warningsystem in terms of area coverage and signalprocessing capacity. This is important as thecurrent system operates with a very simplescheme, that does not allow it to fully utilizethe available seismological information, nordoes it include information on the reliability ofthe early warnings.(3) The self-organizing network is planned toact as an autonomous wireless mesh networkthat works efficiently, even in the event of thefailure of individual nodes (similar to the expe-rimental BerlinRoofNet project (http://sar.infor-

Figure 1: Location of currently used stations for early warning in Istanbul (blue stars) and proposed locations of newstations for expanding the system to the Marmara Area (red stars). Possible locations of the proposed downtown low-cost sensor test network are also shown.

53

matik.hu-berlin.de/research/projects/2005-BerlinRoofNet/berlin_roof_net.htm). To thesenetwork nodes, inexpensive seismic sensordevices are incorporated, resulting in a sensorsystem that will be configured by the appropri-ate hardware and software for application asan earthquake early warning system. Beyondthat, further sensors are to be included thatcan register, for example, building motion orchanges in groundwater characteristics. Thisrequires the appropriate installation of specificanalysis and processing software, as well as sui-table equipment for communication protocolsand services for the cooperative analysis andevaluation of seismic waves and time series ofother parameters. Furthermore, more distantnodes of the sensor system are to be connect-ed by hard wire or over an existing networkinfrastructure (e.g. Internet), thereby not losingits self-organization character (Fig. 2).(4) The EDIM information and mediationsystem must provide appropriate informationto various user groups, e.g. disaster manage-ment, politicians, media, private people, andscientists. For this purpose, a geoinformationinfrastructure will be established that is realizedby components following actual ISO andW3C/OGC standards. Interfaces that link allcomponents are also to be implemented.

State-of-the-ArtSubstantial progress in seismic real-time acqui-sition and communication technologies, asidefrom enhancements of seismic processing soft-

Figure 2: Sensor node network withpotential embeddings

ware, has been made over the past few years.This has paved the way for the design andimplementation of earthquake early warningsystems all over the world (Zschau and Küp-pers, 2003; Kanamori, 2005; Gasparini et al.,2007). Systems are now in operation in Japan,Taiwan, and Mexico (Nakamura, 1989; Wuand Teng, 2002; Espinosa-Aranda et al.,1995), while in Romania and Turkey, systemshave also been recently constructed (Wenzeland Marmureanu, 2006; Erdik et al., 2003).Earthquake early warning systems are effectivetools for disaster mitigation if used for thetriggering and execution of automatisms toprepare vulnerable systems and dangerousprocesses for the imminent danger. Seismicwarnings can be used to slow down high-speed trains to avoid derailments, to close var-ious pipelines, for example gas, to minimizefire hazards, to shut down manufacturingoperations to decrease potential damage toequipment, and to save vital computer infor-mation to inhibit the loss of data. A compila-tion of effective measures in response to suchwarnings is given by Goltz (2002).

InstrumentationOne hundred strong motion accelerometershave been placed in populated areas of Istan-bul, within an area of approximately 50 × 30 km,to constitute a network that will enable rapidshake map and damage assessment after adamaging earthquake. Once triggered by anearthquake, each station will process thestrong-motion data to yield the spectral accel-erations at specific periods and will send theseparameters in the form of SMS messages tothe main data center through available GSMnetwork services. For earthquake early warn-ing information, ten strong-motion stationswere established as close as possible to theMarmara Fault. The continuous on-line datafrom these stations will be used to providenear-real-time warnings for potentially disas-trous earthquakes. In addition, 40 strong-motion recorder units will be placed on criticalengineering structures to augment the alreadyinstrumented structures in Istanbul. All togeth-er, this network and its functions are called the

54

Istanbul Earthquake Rapid Response and EarlyWarning System (IERREWS). The system isdesigned and operated by Bogazici University,with the logistical support of the Governorateof Istanbul, the First Army Headquarters andthe Istanbul Metropolitan Municipality. Contin-uous telemetry of data between these stationsand the main data center is realized with dig-ital spread spectrum radio modem systeminvolving repeater stations. Depending uponthe location of the earthquake (initiation offault rupture) and the recipient facility, thealarm time can be as high as about 8 s. Thus,Istanbul has a basic version of an early warn-ing and rapid response system.The seismological network of the new self-organizing early warning and information sys-tem will be designed to take advantage ofseismological knowledge resulting from manystudies and International projects that havefocused on Istanbul and the Marmara region,especially after the 17 August 1999 Izmitearthquake. New source parameter scalingrelationships (Parolai et al., 2007) and crustalattenuation laws for the area (Bindi et al.,2006) will play a fundamental role, for exam-ple, in the calculation procedures adopted forthe alert and shake maps. On the other hand,for a proper calibration of the attenuation andsource parameter scaling relationships, it is ofprimary importance to also assess the effectsof amplification/deamplification and thelengthening of seismic ground motion due tosurficial geology. For this reason, GFZ andKOERI are carrying out an extensive field cam-paign for the mechanical site characterizationof Istanbul. In particular, single station (about200) and 2D-array (8) measurements havebeen carried out in the western part of themetropolitan area of Istanbul, with the aim ofestimating the S-wave velocity profile for thesedimentary cover and to identify areas whereincreased amplification of seismic groundmotion is to be expected. Therefore, all avail-able seismological and geophysical informa-tion will be used for the optimization of thesingle node analysis (e.g. frequency dependentanalysis for seismic event detection) and forthe improvement of the self-organizing early

warning system as a whole. The new earlywarning system will be designed to comple-ment existing classical seismological monitor-ing networks, but it will focus on the metro-politan area of Istanbul, and the strongground motion and building response.

Sensor developmentIn addition to improvements in the existingcentralized infrastructure, we will also deployand test a decentralized wireless mesh net-work based on inexpensive hardware. Suchnetworks are currently a popular researchtopic in the networking field (Zhang,2006).They aim to be self-organizing in thesense that only a minimum of manual config-uration is required, and that no central admin-istration is necessary. One of the key ideas isthat the services of the network are not pro-vided by a dedicated central infrastructure, butby a collaboration of nodes nearby to therequester; ideally this pattern allows the net-work to be robust against single failures, andto scale well as the number of nodes grows.In effect this enables a network density andreliability which centralized networks can onlyachieve at extremely high costs.In recent years projects like the MIT Roofnetand the Berlin Roofnet (Sombrutzki et al., 2006)have successfully implemented large wirelessmesh networks, primarily to provide free inter-net access in densely populated areas. Theyuse standard WLAN hardware, which utilizeslicense-free spectrum bands for communica-tion. We profit from the hands-on experiencemade in these projects, and apply it to thenew area of geo-sensor networks. To ourknowledge this will be the first study for theuse of self-organizing ad-hoc networks for thepurpose of earthquake monitoring & earlywarning. An evaluation of the real-timebehavior of such monitoring and alerting sys-tems based on self-organizing network infra-structures is almost impossible or too expen-sive without accompanying model investiga-tions. One of the most used engineering toolsis that of an experimental investigation bycomputer simulations. However, computersimulations require and depend on concrete

55

investigation goals. Based on these goals sev-eral models are developed (e.g. in our context:behavioral models of the early-warning sys-tem, transport and alerting communicationprotocols) and suitable simulation frameworksmust be identified (ODEMx library, Gersten-berger, 2003).

Information SystemsThe transfer and processing of heterogeneousspecific information and their integration in aharmonized information system is theapproach part C of the project. Methods ofsemantic interoperability are concerned withthe integration of knowledge from differentapplication domains. In former resarch pro-jects, DELPHI IMM (meanInGs, completed09/2005, www.meanings.de and DeCOVER, inprogress 06/2008, www.decover.info) usedmethods were and developed processes insemantic interoperability between applicationdomains. In order to integrate existing, het-erogeneous data sources automatically it isnecessary to provide a system for the seman-tic interoperability of existing data sources. Amethodic approach is the use of Ontologies[Gruber, 1993]. Ontologies allow the mappingof classes into a hierarchical structure, theclasses are linked to each other by the use ofthesauri, taxonomies and relations. This con-cept provides a high expression strength in aformal description of technical terms. Ontolo-gies are knowledge-based models and realizea kind of description logic for the computer.With this approach it is possible to gain logicalconclusions by feature-based queries (Klienet al., 2004). With the integration of thisapproach in web-services, Geo informationcan be provided to an broad user group.A dynamic geoinformation infrastructure willbe developed in order to disseminate rele-vant information of various sources before,during and after hazardous events for differ-ent target audiences– for the general public, the system will provi-

de aggregated information of latest eventsand personalized guidance. In one centralnode the citizens get localized information intheir native language.

– the local administration and politicians haveaccess for instance to hazard and vulnerabi-lity maps in order to estimate the endanger-ment of their population

– for emergency response coordination by get-ting information during respectively shortafter a hazard for their situation, for instan-ce to coordinate rescue teams and humani-tarian aid.

Based on a spatial data infrastructure (SDI)approach, the following requirements can beaccomplished:– The communication is based on WebServices

supporting international standards of ISO,OGC and W3C. The standardized interfacesenable the interoperability of the compo-nents making the integration of informationfrom different providers into the central por-tal possible: Integration of EDIM internalOGC Web Mapservices (WMS) or OGCWMSs of a local government in the portal(service-to-portal) as well as service-to-servi-ce communication between a OGC WebFeature Service and OGC WMS can beaccomplished.

– In order to support semantic aspects for theinformation system, the use of internationalstandards builds a good and reusable basis.

– The SDI concept provides a distributed net-work of components connected via theInternet based on the HTTP protocol. Thedistribution of services creates a basis fordecentralization and reducing the vulnerabi-lity of the entire system.

Each service is a module in the infrastructurewith a different task: While sensors are col-lecting data like accelerations at the one end,the information are disseminated via a stand-ardized interface of OGC sensor observationservice (SOS) to the central portal for theemergency crew. A Web Feature Service (WFS)stores vector data of roads and informationabout current road conditions. Users that wantto know if a road is blocked after a hazardousevent, can log into the portal and check a mapwith that information rendered by a WMS.Depending on their rights they have access tomore confidential information controlled bythe owsPproxy component. The portal is the

56

central information node where user’s gettheir localized and personalized information.By using standardized interfaces, services fromother networks can be hooked in transparent-ly to the user. In between the portal and ser-vice a data processing takes place in order toaggregate data and make information of dif-ferent sources semantically comparable.

First ResultsThe existing data transmission is via 2.4 GHzSpread Spectrum Radio Modems and relay sta-tions. Currently, a second data transmissionavenue is implemented through the use ofsatellites, involving the installation of DirecwaySatellite dishes at each station to transmit thedata to the main center. This produces a timedelay of about 0.5s. Dexar Company in Istan-bul (affiliated with U.S. Hughes Network Sys-tems) will provide this service. The future plan

involves adding a third transmission system viaADSL. To provide total coverage in the Mar-mara Region (Fig. 1) the current 10-stationnetwork (blue stars in the attached figure)needs to be expanded with an additional 10stations (red stars in the attached figure).

Early Warning MethodologyIn cooperation with KOERI, a novel scheme forextracting early warning information has beendeveloped within the framework of a PhD pro-ject (Böse, 2006) jointly supervised by TH (Prof.Friedemann Wenzel) and KOERI (Prof. MustafaErdik). Some of the results have been pub-lished in Böse et al. (2005) and Böse et al.(2006). Fig. 3 shows an example of an alertmap for Istanbul for the case of a M = 6.5earthquake in the eastern part of the MarmaraFault. The left panel indicates the distributionof PGA after the earthquake: the right panels

Figure 3: Alert Map for Istanbul

57

show the predicted alert maps after 3.5 and4.2 seconds. After 4.2 seconds the map comesfairly close to the final ground motion, so thatafter this time the regional ground motion isfairly well predicted.The actual warning time for Istanbul for thisearthquake is 15.8 seconds. The partners inKOERI developed the existing warning andinformation system, which includes the instal-lation of sensors, real-time communication,shake-map generation and earthquake-lossestimation technology (Erdik et al., 2003). Theneural net methodology PreSEIS (Böse, 2006)for earthquake early warning has been cali-brated with real data from Southern Califor-nia. The dataset consists of 69 shallow localearthquakes with a moment magnitude rang-ing between 1.96 and 7.1. The data comefrom broadband (20 or 40 Hz) or high broad-band (80 or 100 Hz), high gain channels (3-component). Although the data have beenrecorded at a total of 177 stations of theSouthern California Seismic Network (SCSN),only certain combinations of single stationshave been tested so far. As an input for theneural nets, the envelopes of the waveformsdefined by Cua (2004) are used. The envelopeis obtained by taking the maximum absoluteamplitude of the recorded ground motiontime history over a 1-second time window.Due to the fact that not all of the stationsrecorded each earthquake, the missing recordswere replaced by synthetic envelopes, calcu-

lated by envelope attenuation relationshipsdeveloped by Cua (2004).

Sensors and NetworksThe model-driven development (MDD) initia-tive of the OMG proposes a transfer from plat-form independent models to platform depen-dent models. We will use this approach torealize the complex task of simulating the real-time behavior of the ad-hoc mesh sensor net-work for EEW.Therefore we designed and partly implement-ed a prototyping infrastructure with the fol-lowing characteristics:– A repository with customizable and executa-

ble model components, which enables theefficient configuration of executable EEWSmodels (Experiment Management System).

– Model components in the repositorywhich represent software components ofspecific seismic, alerting and distributedservices should be evaluated by simula-tion experiments.

From these models corresponding softwarecomponents can be generated for the EEWSand deployed on the target platform. Thisensures that the same software componentwhich was evaluated by simulation runs onthe target platform.We designed a component-based architectureto be deployed on every node (see Figure 4).It consists of both functional as well as tech-nical software components. Functional soft-

Figure 4: Software Architecture of a Node

58

ware components, like the analysis of sensordata, reflect the core services the self-orga-nizing network has to provide. Functionalcomponents are first implemented and testedwithin the simulation part of the prototypinginfrastructure. In contrast, the technical com-ponents are driven by the needs and restric-tions of the underlying hardware infrastruc-ture, and are therefore developed directly forthe platform. Components already imple-mented include the retrieval of data withinthe node (Data Sensing, from the sensors aswell as from a file), and the supply of data(Raw Data Archiving) this through the wirelessnetwork. This allowed us to conduct firstexperiments in a small in-house test bed withreal hardware.The component Map Parameter Calculationcomputes the peak ground acceleration (PGA)necessary for ground shake maps as well asfor alert maps. The component CollaborativeEarly Warning Management decides how todeal with incoming alarm messages fromother nodes (and work in progress).A first version of a low-cost sensor system(ADXL203) for three-component accelerationand/or velocity (Figure 5a) measurements hasbeen developed by GFZ Furthermore, takinginto account the low-cost margin of theplanned system that makes prohibitive the useof expensive commercial AD-converters (AD-c),a 4-channel AD-c was developed by the GFZ(Figure 5b). These converters are designed forsample rates of up to 400 Hz with a signal

bandwidth from 0 Hz to 80 Hz. Moreover, alow-pass a-casual filter system with a hard-ware adjustable corner frequency before digi-tization to be integrated into each sensor hasbeen developed.In cooperation with the project partner HU, asuitable low-cost hardware platform has beenidentified for reading and handling the datastreams from the AD-c board. The hardwareconfiguration for the WLAN computer is basedon the use of a single board computer, the PCEngines WRAP.2E board (Figure 5c), which isequipped with two 802.11a/b/g WiFi cards(Atheros chipset), a 1GByte CFlash card andtwo USB ports, of which one is used to attachthe AD-c converter developed by the GFZ.The communication system between sensors isdeveloped taking advantage of the experienceof the operating seismological real-time sys-tem for worldwide earthquake monitoringdeveloped by GFZ within the context of theGEOFON Project. In fact, the SeedLink soft-ware installed on the WRAP computers allowsdata exchange for a first prototype seismolo-gical network, and a plug-in has been writtento handle the data streams from the AD-c.The new seismic sensors, properly equippedwith a integrated USB port for data transportto the mesh point of the network and GPSinstruments for time synchronization and loca-tion, will be tested at the GFZ during a pre-liminary field experiment. In particular, theusage of a limited prototype network on theTelegrafenberg Albert Einstein Science Park,

a) b) c)

Figure 5: a) Three accelerometers (ADXL203 from Analog Device) with an analog filter form the 3 component low-costseismic sensor. b) A new 24 bit low cost AD converter for 4 channels. GPS will be integrated for timing and locationdetermination. c) WRAP computers handle the data streams from the sensor board. Data exchange with neighboringsensors will be realized with WLAN.

59

both inside and outside buildings, will providethe opportunity to evaluate the sensitivity ofthe low-cost 3 component accelerometers, thepower consumption of the individual compo-nents and of the complete system, and thegeneral reliability and stability of the system,paying particular attention to both the per-formance of the individual nodes and com-munication between nodes. At present, basicsoftware for filtering, STA/LTA triggering, andevent detection and discrimination has beendeveloped. Therefore, during the field experi-ment, the software devoted to run on eachsingle node will be tested.

Information SystemsIn EDIM a number of application scenarios arebeing developed and validated regardeing thesetup of a geodata-infrastructure, for the inte-gration of geodata and geo-services as well asfor an intelligent search of geo-informationapplication scenarios are being developed andvalidated. During the development of thesemantic interoperability of geo data, the fol-lowing services are focused on: Support of theinformation provider, support of the informa-tion inquirers, mediation services betweeninformation provider and information inquirersand an automatic process to perform qualitycontrol by the use of Service Level Agreements(SLA). Concrete results of this approach areseen in language independent risk estimation,real-time warnings, a fast disaster evaluationand support of the preparation of effectivesupporting measures. The following applica-tion scenarios are seen as vital regarding userrequirements. Geology and earthquake cata-logue – this scenario describes the integrationof existing geoscientific, earthquake and dis-aster information in a knowledge-basedmodel. A second scenario shows the integra-tion of land use information, infrastructuredata and possibly socio-economic data stets.Relevant classification catalogue for earlywarning and disaster systems can be producedby ontologies and similarity measures makedetermined warnings and assistance possible.A further application is to be seen in the useof Earth Observatory satellites. In this case it

must be examined whether a derived classifi-cation catalogue can be applied for a guideevaluation shortly after the case of disaster.For the entire system Open Source compo-nents will be used:– For the database backend PostgreSQL with

the spatial extension PostGIS will be used.– For all OGC WebServices like WMS, WFS and

SOS deegree services will be used.– The security components are provided out of

the degree project as well.– For the central information node, a portal

framework (Apache Jetspeed) will be used.The clients for each OGC WebService will berealized as portlets conforming to the JSR-168 standard.

Currently, a new geometry and feature modelfor deegree3 are designed to support theupcoming ISO and OGC standards. A newconcept for the new client framework sup-porting both web and desktop clients is devel-oped. For the first prototype deegree2 com-ponents will be used presenting the currentpossibilities. Step-by-step, they get updated bydeegree3 components for using the neweststandards. The interaction of all (already inuse) components will be accomplished by sup-porting the according standards.

SummaryThe EDIM consortium has the potential andstrategy to address key issues in further devel-oping the Istanbul Earthquake Rapid Responseand Early Warning System (IERREWS). We willbridge the gap between single station alarmsystems and those requiring records from anentire network, using the Artificial Neural Net-work methodology (Böse et al., 2005) thatallows an alarm to be issued from one station,but which will be upgraded with each addi-tional available record. In cooperation withGFZ and HU, we will develop a fundamentallynew monitoring and communication system,based on low-cost, self-organizing recorderswith a high potential for applications in manyplaces of the world.We will address the rapid inversion of ourceparameters as critical input for city-wide dam-age predictions. We will study neural net

60

methods, specifically adapted for the MarmaraRegion, that do not require velocity models,but need the incorporation of experience fromprevious events and simulations. Algorithms tobe developed will consider the degree of reli-ability of the information about forthcomingground shaking and damage. For example, theneural net methods developed at TH allow theenhancement of the reliability of the earlywarning with evolving time.Linking ground motion information withdamage modelling (KOERILOSS) aims at therapid quantification of economic damage tothe region, so that socio-economic implica-tions of an earthquake can be understoodimmediately after the event. Self-organizingnetworks can deliver much more detailedinformation about the functionality of specif-ic structures and critical facilities. Data fromthe warning system, including the self-orga-nizing sensors, and further spatial informa-tion will be provided by an interoperableinformation infrastructure.

ReferencesBindi, D., Parolai, S., Grosser, H., Milkereit, C.,and Karakisa, S., 2006. Crustal attenuationcharacteristics in northwestern Turkey in therange from 1 to 10 Hz. Bull. Seism. Soc. Am.96(1): 200–214.

Böse, M., Erdik, M. and Wenzel, F., 2005:Earthquake Early Warning – Real-time predic-tion of ground motion from the first secondsof seismic recordings. Proc. Volume of the Int.Conference on 250th Anniversary of the 1755Lisbon Earthquake, 185–187, 1–4 Nov. 2005.

Böse, M. Erdik and F. Wenzel, 2006: A NewApproach for Earthquake Early Warning. In:J. Zschau and P. Gasperini (eds.): Seismic EarlyWarning Systems, Springer Verlag, in press.

Böse, M. 2006: Earthquake Early Warning forIstanbul using Artificial Neural Networks. PhDthesis, University of Karlsruhe, Germany.

Cua, G., 2004: Creating the Virtual Seismolo-gist: developments in ground motion charac-

terization and seismic early warning. PhD the-sis, California Institute of Technology, USA.

Erdik, M., Fahjan, Y., Özel, O., Alcik, H., Mert,A., Gul, M., 2003: Istanbul Earthquake RapidResponse and the Early Warning System. Bul-letin of Earthquake Engineering, 1, 157–163.

Espinosa-Aranda, J., Jimenez, A., Ibarrola, G.,Alcantar, F., Aguilar, A., Inostroza, M., andMaldonado, S., 1995: Mexico City SeismicAlert System. Seismological Research Letters,66, 6, 42–53.

Gasparini, P., Manfredi, G. and Zschau, J. (eds)Earthquake Early Warning Systems, 349 pages,Springer, 2007.

Gerstenberger, R., 2003: ODEMx: Neue Lösun-gen für die Realisierung von C++-Bibliothekenzur Prozesssimulation. Diplomarbeit Hum-boldt-Universität zu Berlin.

Goltz, J.D., 2002: Introducing earthquake earlywarning in California: A summary of social sci-ence and public policy issues,. Caltech Seis-mological Laboratory, Disaster Assistance Divi-sion, A report to OES and the OperationalAreas.

Gruber, T.R., 1993: A Translation Approach toPortable Ontology Specifications, In Knowl-edge Akquisition Vol.5, No.2, 1993.

Kanamori, H., 2005: Real-time seismology andearthquake damage mitigation. AnnualReviews of Earth and Planetary Sciences, 33,5.1–5.20.

Klien, E., Lutz, M., Einspanier, U., & Hübner, S.,2004: An Architecture for Ontology-Based Dis-covery and Retrieval of Geographic Informa-tion. Proceedings of the 7th Conference onGeographic Information Science (AGILE 2004)

Nakamura, Y., 1989: Earthquake alarm systemfor Japan Railways. Japanese Railway Engi-neering, 8, 4, 3–7.

61

Nakamura, Y., 2004: UrEDAS, Urgent Earth-quake Detection and Alarm System, Now andFuture. Proceedings 13th World Conferenceon Earthquake Engineering, Vancouver,August 1–6, 2004.

Parolai, S., Bindi, D., Durukal, E., Grosser, H.,and Milkereit, C., 2007. Short Note SourceParameters and Seismic Moment–MagnitudeScaling for Northwestern Turkey Bull. Seism.Soc. Am. 97(2): 655–660.

Sombrutzki, R., Zubow, A., Kurth, M., andRedlich, J.-P. Self-Organization in CommunityMesh Networks – The Berlin RoofNet, FirstWorkshop on Operator-Assisted (WirelessMesh) Community Networks, pp. 1–11, 2006

Wenzel, F. and Marmureanu, G., 2006: Earth-quake Information Systems, in Pageoph Topi-cal Issue, Proc. Volume 22nd Int. Tsunami Sym-posium, Chania, Greece, 27–29 June, 2005, inpress.

Wu, Y.-M. and Teng, T.-l., 2002: A virtual sub-network approach to earthquake early warn-ing. Bulletin of the Seismological Society ofAmerica, 92, 5, 2008–2018.

Zschau, J., N. Küppers (eds), Early WarningSystems for Natural Disaster Reduction, 834pages, Springer, 2003.

Zhang, Y., Luo, J., and Hu, H. (eds) WirelessMesh Networking: Architectures, Protocolsand Standards, 592 pages, Auerbach Publica-tions, 2006

62

Numerical Last-Mile Tsunami Early Warningand Evacuation Information System

1. BackgroundSeventeen out of twenty of the most disas-trous natural hazards, since 1950, haveoccurred in the last 10 years. Extreme envi-ronmental events grow in frequency and mag-nitude. Hence, the number of human lossessince the 1950’s has now reached 1.7 million;and the economic damage exceeding 1.4 Bil-lion US$.1 Most of the hazards affect morethan just one region, the most prominentexample being the devastating tsunami of 26December 2004 in the Indian Ocean. Thisaccording to the current estimates killed morethan 220,000, made more than 1 million peo-ple homeless, and left many thousands with-out any basis of existence. The estimated eco-nomic costs (10 billion US$) of the tsunami iscomparatively low. However, only 20% of thissum has been registered as insured loss.2 Thisclearly indicates that the aftermath of theseaquake to the west of Sumatra also consistsof certain unforeseen dimensions. Therefore asignificant shift in risk perception of anextreme event must follow. The aftermath ofthe tsunami, with all its consequences hasalready affected the development politics andthe international, multi-sectoral research com-munities. The overarching goal of these initia-tives is to improve disaster prevention by pre-paredness methodologies, including public

and political awareness of the residual risk andaddressing the global hazard resilience. Thecurrent multi-disciplinary project »Last-Mile –Evacuation« attempts to incorporate all ofthese guiding principles and jointly developswith local partners the necessary evacuationrecommendations by scenarios of a tsunamiearly warning and evacuation information sys-tem. The reasons for the disastrous impacts ofhazards, especially in developing countries, aremanifold: rapid population growth, growingsocio-economic segregation in most parts ofthe population and the extreme growth ofurban agglomerations in coastal areas are themajor factors. This automatically creates adilemma, where growing utilisation and settle-ments pressures in coastal areas have to bebalanced against preserving a natural land-scape. A natural coastline and its native greenbelts play an important role in protecting aregion from both tsunamis and storm surges.3

The narrow coastal zone up to 100 kilometresinland represents only 6 per cent of the glob-al land area. However, nearly 40% of theworld’s population lives in this small strip, andthis tendency is growing constantly. On top ofthis, two thirds of all major cities countingmore than 2.5 millions citizens can be foundin exactly this coastal zone.4 The vulnerabilityof the people living on the coastal zone will

Birkmann (1), Dech (2), Hirzinger (3), Klein (4), Klüpfel (5), Lehmann (3), Mott (6), Nagel (7),

Schlurmann (8), Setiadi (1), Siegert (6), Strunz (3)

(1) United Nations University, Institute for Environment and Human Security

(2) University of Würzburg

(3) German Aerospace Center (DLR)

(4) University of Bonn

(5) TraffGo HT GmbH

(6) Remote Sensing Solutions GmbH (RSS)

(7) Technical University Berlin

(8) Leibniz University Hannover

63

grow constantly. It seems obvious that onlysustainable development will deliver solutionsto decrease disaster risks in coastal zones.Therefore the sustainability’s basic principlesand criteria for protection and use mustinclude economic efficiency; and economicintegrity while ensuring equitableness.5 Ingeneral those attempts consider the majorconcepts and principles of subsidiary and/orreversibility; this project in the context of earlywarning systems (EWS) in earth managementincludes prevention as the main solution ofdisaster risk reduction. The danger of naturaldisasters lies in the complexity of hazards,human vulnerability, coping capacity, re-silience and the given environmental realitieslike water, drinking water, soil and agriculturein coastal zones. A disaster is more likely tooccur, if the community is not informed,aware or perceived of the existing risks. How-ever, facing the ever growing frequency andmagnitude of natural events shows how cru-cial disaster preparation have become. Reduc-ing human vulnerability is the only way tominimise the risk of a natural disaster. Besidesactivities to raise public awareness, the devel-opment of early warning systems is indispen-sable.6 The lives of many people and the levelof economic loss depend strongly on thosesystems. The overarching ambition of »Last-Mile – Evacuation« within the context of earlywarning systems in earth management is todevelop and test a tsunami early warningand evacuation information system, whichwill be exemplarily implemented as a proto-type application in the city of Padang, WestSumatra, Indonesia.The term last-mile itself characterizes the abil-ity to secure the proper execution of an earlywarning chain in all its facets and institutions,i.e. from intergovernmental commissions overnational bodies down to coastal communitiesand the individual himself.7 The image last-mile was first drawn by defining the so-calledcrucial link after the devastating earthquake inBam, Islamic Republic of Iran, in 2003. Itdescribes the interface in between moderncommunication technologies and the pro-posed recipients of the information – the ten-

tatively affected people. Any design of an earlywarning system has to recognized and subse-quently embed local structures, actors andcapacities to create an end-to-end EWS. Localauthorities and affected people have to bemade aware and adequately prepared of nat-ural hazards in order to anticipate issuedwarning dossiers in a proper way and reactadequately. Thus, it is likewise a primary goalof this project to raise awareness and developappropriate preparedness mechanism byachieving a numerical last-mile tsunami earlywarning and evacuation information systemwhich delivers concrete evacuation recom-mendations for tsunami inundation scenariosand methodologies for optimized evacuationcircumstances with the local authorities andscientist in a joint research approach.Moreover, it is essential to recognize that thelocal government of a district or city plays themost significant role when developing and ini-tiating to execute research projects in Indone-sia due to the fact the decentralization anddemocratization processes are still in progressto advance in the country. It is evident that dis-tricts (Kabupaten) and cities (Kota) permanent-ly receive more and more competencies andresponsibilities from the Indonesian govern-ment; especially in relation to disaster man-agement and prevention policies. By constitu-tion the Lord Mayor has the authority of mak-ing decisions in disastrous events, e.g. whethera city or coastal stretch is being evacuated ornot. The long-term and sustainable distributionof resources and civil protection of citizens ina city is steered by the local authorities. In fact,the stringent decentralization process within acountry like Indonesia is beneficial for cus-tomized local disaster management schemesand likewise advantageous for an anticipatedrapid response to tsunami warning dossiers interms of evacuation. Therefore, the projectmanagement of »Last-Mile – Evacuation« hasto elucidate whether the local authorities andcapacities are yet defined by an efficient com-bination of effective, interdisciplinary, integrat-ed and wide-ranging managing institutions fordisaster management plans in the city ofPadang, and, should initiate guidelines to set

64

an institution like this into effect for the fullusage of the information tool being developedin this project.

2. Description of the Research ProjectThis research project develops a numerical lastmile tsunami early warning and evacuation infor-mation system (acronym: »Last-mile – Evacua-tion«; sponsorship code: 03G0643A-E) on thebasis of detailed earth observation data andtechniques as well as unsteady, hydraulicnumerical modelling of small-scale floodingand inundation dynamics of the tsunamiincluding evacuation simulations in the urbancoastal hinterland for the city of Padang, WestSumatra, Indonesia. It is well-documented thatSumatra’s third largest city with almost onemillion8 inhabitants is located directly on thecoast and partially sited beneath the sea level,and thus, is located in a zone of extreme riskdue to severe earthquakes and potential trig-gered tsunamis. »Last-Mile – Evacuation«takes the inundation dynamics into accountand additionally assesses the physical-technicalsusceptibility and the socio-economic vulnera-bility of the population with the objective tomitigate human and material losses due topossible tsunamis. By means of discrete multi-agent techniques risk-based, time- and site-dependent forecasts of the evacuation behav-

iour of the population and the flow of trafficin large parts of the road system in the urbancoastal strip are simulated and concurrentlylinked with the other components. This projectis being developed in close cooperation withthe local authorities of Padang, the Engineer-ing Faculty of the Andalas University in Padangas well as the Indonesian research institutions.Thus, by means of modelling tsunami inunda-tion scenarios as well as performance of evac-uation processes a crucial disaster prepared-ness measure is made available. Developingand implementing a suitable software applica-tion, i.e. an information system, disaster miti-gation methodologies are created and opti-mized and can be indirectly steered prior andduring a catastrophic natural extreme events.The time framework allocated for »Last-Mile –Evacuation« is three years.

2.1. Scientific and technological objectivesIn general, an adequate disaster preparednessscheme of any early warning system generallyconsists of four or five levels, respectively.Those stages, which are consecutively applied,build the so-called early warning chain. Onlythe interaction of all stages can assure aneffective early warning system (e.g. in the caseof a tsunami):

Figure 1: One of Padang main roads located along the coastline

65

This research project mainly addresses levelfour, but also taking into consideration thelevel five of an early warning system in the cityof Padang. To deliver recommendations andmodel the evacuation behaviour on this levelof an effective early warning chain, the projectaims for deriving tsunami inundation dynamicsand evacuation simulations due to extremenatural events like tsunamis on the localcoasts. This is done by building an informationsystem for subsequent disaster mitigationmethodologies that can be indirectly steeredprior and during a catastrophic natural ex-treme events.The determining aspects of any numerical lastmile tsunami early warning system in coastalcommunities (level 5) are so far being neglect-ed from a scientific point of view. In front ofthis background the major task of futuredevelopments must be focussed on appliedtsunami research themes as well as in thegeneral understanding of the reciprocal rela-tion betweenTsunami – Coast (urban hinterland) – People’sbehaviour (reaction)in case of concrete tsunami warning, or e.g.likewise in case of storm surge warning. Theoverarching objective is to deliver precise pre-paration measures regarding clear warning andevacuation schemes regarding the time of theday and the day of the week specific depen-dencies in near shore, urban agglomerations.The project funded by the Federal Ministry ofEducation and Research (BMBF) with the title»Implementation of core elements of a tsuna-mi early warning system in Indian Ocean withcollaboration of Indonesia and other partners– GITEWS« (submission number: 03TSU01)

aims to develop an effective and sustainableearly warning system with Indonesia and othercountries in the Indian Ocean Rim. This majorproject of the German Helmholtz Institutionsand other research centres mainly concen-trates on monitoring systems and innovativetechnologies (levels 1 and 2). A possibleseaquake in the region of Sunda Arc possiblycausing a tsunami, might reach the WestSumatran coast of Indonesia within 18–20minutes. This critical time mark defines thedesign and the evaluation of data of thetsunami early warning system in Indonesia.9

Therefore the tsunami early warning systemcan only make sense and a meaningful differ-ence if the issued warning dossier (or all-cleardossier) reaches the affected region within10–12 minutes to initiate a proper evacuationin the coastal region.»Last-Mile – Evacuation« originates itself uponsome of the information and evaluations thatare being developed or in the process of beinggenerated in GITEWS. In this respect certainwork packages are directly related to thisresearch project since the tsunami risk assess-ment studies in GITEWS are being conductedfor the whole Indonesian coastal stretches(approximately 4000 km) bordering the IndianOcean (Sumatra, Java, Bali, etc.). Thus, it is agiven fact that coarse resolution analysis isconducted in GITEWS, which concentrates onresearch on deep-water tsunami propagationin the whole Indian Ocean as well as to pa-rameterize the interaction between tsunamigeneration and seismic strokes. The localtsunami run-up is being analyzed similarly, buthydro numerical simulations stem from ameso-scale numerical grid size and irregular

Table 1: Levels of an effective early warning chain

66

cell structure (100 × 100 m, occasionally finerresolved: 50 × 50 m).The hydro numerical models used in commonapproaches do neither incorporate local infra-structures nor do they consider street net-works or urban waterways in the respectivecities. The simulations make use of constantcoastal slopes representing the near shoreregion and the urban hinterland withoutdetermining infrastructures. To establish quan-titative information about the local run-upheights onshore, land use and land cover dataare often parameterized to imply the charac-teristic roughness and thus make use of fric-tion coefficients for the model. Neither numer-ical evacuation simulations are envisaged incommon research projects nor an informationsystem relying on detailed, time-dependentinundation dynamics and evacuation recom-mendation are produced. Thus, it can be eas-ily argued that »Last-Mile – Evacuation«apparently erects its scientific basis upon pre-vious scientific studies, but goes far beyondthe expected outcome. In summary, »Last-Mile– Evacuation« defines itself as an independentresearch project. It is neither seen as a simul-taneous scientific challenging approach nordoes it produce redundant results for valida-tion purposes.In reference to table 1, »Last-Mile – Evacua-tion« completes the early warning chain andexemplarily shows the mechanisms of evacu-ation efforts and the dynamics of tsunamiinundation in an urban area in WesternSumatra. Thus, the project addresses the cru-cial questions in how – after a verified tsuna-mi scenario alignment and thereby validateddissemination – the initiation for evacuation isaffected and the subsequent chronology ofthis procedure is feasible at all. What are theconsequences of the impact and reactions ofthe people in Padang which will lead to theformulation of evacuation recommendationor technical preparedness measures (i.e. shel-ters). It is reasonable to allocate the project’smethodological approach to other coastalcities in the Indian Ocean Rim by substitutingmerely the boundaries conditions and provid-ing a best-practice solution.

2.2. Overall project objectivesThe key objective of this research project is toachieve and fulfil the final level of the tsuna-mi early warning chain in the city of Padang.Answers as well as concepts to the followingcentral aspects are determined:– In what respect is – after localization of a

tsunamigenic earthquake, after detectingthe triggered tsunami in the deep ocean andafter verified scenario alignment and therebyvalidated dissemination in the crisis andinformation centre in Jakarta – the initiatedrequest for evacuation and the chronologyof this procedural instruction including theinherent physical-technical susceptibility andthe socio-economic vulnerability of thepopulation in the coastal region integratingthe respective daytime and weekday feasibleat all?

– What kind of inundation dynamics charac-terizes the tsunami in the city of Padang andwhich consequences can be derived for theoptimization of evacuation schemes andstandard operational procedures? What arethe exact time frames due to tsunami inun-dations in Padang’s coastal districts to suc-cessfully organize evacuation routines?

– Which bottlenecks arise during the processof evacuation? What time of the day andwhat day of the week specific dependenciesemerge during the evacuation? Where docritical infrastructures subsist and how istheir individual evacuation being optimized?How is the general traffic situation within anacute tsunami threat after the onset of eva-cuation? How might these blocked obstruc-tions and bottlenecks be resolved by now?

– How can the vulnerability of the populationand those of the critical infrastructures beexemplified and measured? When andwhere are hotspots of vulnerability andwhere is special assistance in the case of anevacuation needed?

– Which scenario-specific decisions andrecommendations, e.g. vertical evacuation,tsunami shelters, etc., have to be met tech-nically and within future administrative spa-tial planning processes in order to minimizethe tsunami disaster risk in Padang, i.e. to

67

reduce the susceptibility of Padang’s coastand the vulnerability of its people?

On the whole »Last-Mile – Evacuation« leadsto insights about the detailed flow dynamicsof tsunamis as well as it pinpoints to opti-mized early warning and evacuation mecha-nisms in the city of Padang which can findoperation to municipal and spatial planningefforts in an attempt to deal with an integrat-ed coastal zone management – here inPadang.A site-specific software application using aGraphical User Interface (GUI) is defined as thephysical end product of the project. This sys-tem assembles scenarios, data sets and analy-sis results of the four thematic subcomponentsof the project and inherently generates addi-tional, coupled information for decision-mak-ers and local authorities. It is feasible to assignthe project’s methodological approach toother coastal cities in the Indian Ocean Rim.This system is being developed in close coop-eration with local authorities and researchinstitutions as a participative approach andis finally recommended as a supplementarytool for the local decision-makers since it inte-grates an efficient and effective collection,administration, processing and distribution ofdata as well as it generates and hosts analysisresults of early warning and evacuationprocesses in the city of Padang. It thus alsoincorporates approaches for the developmentof adaptation and mitigation strategies toreduce disaster risks in the region; especiallyby means of Capacity Building and Stakehold-er Dialogue to generate knowledge and fur-ther insights about tsunami threats and appro-priate evacuation recommendations for thepeople of the region.The end products of the project, i.e. the evac-uation recommendations & information sys-tem, are gradually established. The once gen-erated rudimentary version of the applicationis recommended to the local authorities as abasic tool for daily planning processes. Itembeds an efficient and effective collection,administration, processing and distribution ofdata and the inherent information from calcu-lations concerning early warning and evacu-

ation. The information system’s architecture isbuilt as an OpenSource code to enableautonomous upgrades by the local authorities.Interfaces and standards are defined to feedupdates into the system, e.g. field survey dataor newly derived census data. On the whole,the project demands for and yet, incorporatesa interdisciplinary research approach (social sci-ences, economy, engineering and natural sci-ences) together with a strong linkage to localauthorities, decision-makers and the people ofthe city of Padang.

2.3. Particular scientific and technologicalwork loadThe work load towards the research projectobjectives is divided in five working packages,based on field of expertise of each workingpartner, namely socio-economic vulnerabilityassessment (WP 1000), inundation scenarios,flow analysis (WP 2000), geodatabase, infor-mation system and vulnerability assessment(WP 3000), evacuation analysis and pedestriantraffic optimization (WP 4000) and highly-resolved 3D-model of Padang (WP 5000) (SeeFigure 6 in the Chapter 3). On the basis ofsatellite images and other data sources,advanced highly resolved information includ-ing topography, infrastructure, populationdensity and settlement structures are provided.These tasks are compiled by WP 3000. Col-lected data serve as starting point for furtheranalyses. In cooperation, WP 1000 focuses onthe analysis of the susceptibility and copingcapacity of the inhabitants in certain ranges ofthe coastal stretches in Padang to exemplifyand measure vulnerability in a broader sense.This is done in a holistic endeavour to resolvethe hotspots of vulnerability and to ascertainwhere in case of an evacuation special assis-tance is needed. These widespread evaluationsare linked to the coupled information systemthat is being developed upon all assembleddata by WP 3000, so that real time prognosesafter a tsunami battered the shores can givefirst estimates of tentative losses and potentialdamages. Thus, essential information for adhoc disaster management activities can bedetermined effectively.

68

WP 1000 concentrates on research aspectshow to specify vulnerability indicators withregards to exposed and susceptible elements(people/different social groups and their socio-economic structures), as well as evacuationcapability, perception and behaviour. It com-prises the quantitative collection of suitableindicators of Padang’s population throughlocal municipal statistics (e.g. Census), as wellas additional surveys. Moreover, indicators oncritical infrastructures will be collected in col-laboration with WP 3000, which focuses toconduct an extensive physical vulnerabilityassessment. The use and unification of differ-ent data sources are essential for the decisivedevelopment of insightful vulnerability indica-tors. It provides a deeper understanding andcharacterisation of Padang’s districts to comeup with evacuation information in case of aheightened tsunami threat. Data for these par-ticular analyses are collected and processed bylocal researchers and students from theAndalas University in Padang.This holistic approach on vulnerability assess-ment aims preliminary on small-scale andhousehold level investigations. It focuses on athorough assessment of the vulnerability ofthe society and population, economics andenvironment by natural disasters in particularby tsunamis or e.g. storm surges, whichrequire special attention when establishing aninformation system associating evacuationroutines. In addition this vulnerability analysisintends to elucidate the characterisation and

quantification of coping capacities and its spa-tial distribution in the city of Padang to deter-mine existing institutions for disaster controland management (fire-brigade, police) andtheir situation as well as appropriate emer-gency shelters for evacuation. This latteraspect is linked with a particular componentfrom GITEWS. The work package CapacityBuilding also strengthens local governmentsand organizations to develop appropriatewarning and disaster preparedness mecha-nisms; mostly on the basis of drills and train-ings in the field. This component is carried outby the German Technical Corporation (GTZ)and is intended to put three pilot regions inIndonesia into operation – also in Padang.Detailed hazard maps due to tsunamis or e.g.due to storm surges for coastal areas on thebasis of one- and two-dimensional unsteadyhydro numerical simulations are done by WP2000. It develops tsunami inundation scenar-ios in »Last-Mile – Evacuation«, which areestablished on shallow water bathymetry andurban topography of the coastal hinterland.The requirements in terms of precision anddetail for a digital elevation model (DEM) areextremely high. It is agreed with the Indone-sian researchers from the line-agencies fromBAKOSURTANAL, BPPT and LIPI to jointlyextend the simple digital elevation model forthe city of Padang in a sustainable way. WP2000 and 3000 support the Indonesian coun-terparts in upgrading that DEM. Spatially high-ly dissolved data of the bathymetry are indis-

Figure 2: Physical vulnerability of builtinfrastructure in Padang (Source: DLR)

69

pensable due to the major impact of sub-merged small-scale structures, e.g. coral reefs,sea canyons, headlands, sandbars or fluvialnear shore sedimentations, regarding thehydrodynamic simulation of the wave trans-formation once the tsunami reaches theshores. It is known from Indonesian re-searchers that coarse bathymetry data areavailable, but further measuring campaignsare being conducted by WP 2000.Once the DEM is extended, WP 2000 conductsmicro scale hydro numerical simulation of nearshore tsunami wave dynamics off the coast ofPadang and links these simulations with thebore propagation of the incoming wateronshore on different numerical scales. Ahybrid approach is chosen that defines itsboundary conditions from the variation ofwater levels and near shore currents from apre-selected hydro numerical model of thenear shore coastal wave field. In the transitionzone between the beach and the urban areaan efficient, and rather rapid, one dimension-al hydro numerical model is accordingly set-upand integrates significant open paths andwaterways (streets, rivers) as well as the capa-bilities of retention areas (open fields, parks,markets). This model is extracted from thehighly resolved DEM (WP 3000 and WP 5000).It is then subsequently provided to WP 4000for further usage in their respective researchprogrammes. In the meantime, this basicmodel is expanded to shape a two-dimension-al adaptation that is feasible to cover specific,critical infrastructures (hospitals, schools orother public buildings) in detail. By means ofnumerical coupling this enhanced model isintegrated in the existing one-dimensional rep-resentation of the city of Padang and whollylinked with WP 4000 which subsequently aimto model how evacuation procedures in spe-cific infrastructures and streets evolve basedon the established hydro numerical model pro-vided by WP 2000. As a further result for sub-ordinate implementations, e.g. the detailedsurge flow dynamics past selected infrastruc-tures, the parameters water level, flow rateand direction at any place of the city for theentire period of the tsunami inundation are

made available in the urban area. The preci-sion and detail of any pre-calculated scenariois provided in spatial domains in the order of1.0 meter, while temporal resolutions areabout 5 seconds. These highly resolved inun-dation mechanisms in the city of Padang thenenable to estimate relevant potential damagesand provide decision-support parameter togenerate unambiguous vulnerability assess-ments, and consequently derive optimizedevacuation procedures. This directly establishesthe linkage to the subproject of WP 4000.For the development and examination of evac-uation concepts, WP 4000 use discrete multi-agent-models which rely on the substantialaspects of the evacuation process itself. Thus,extremely fast and most efficient numericalsimulations are feasible. WP 4100 concen-trates on evacuations within buildings and typ-ical infrastructure (blocks) in the city ofPadang. Design sketches of any structure pro-vide the basis for these computations. Bymeans of thorough data collection, WP 1000delivers typical daytime and weekday informa-tion about the habits and behaviours of thepeople and different social groups to gain adeeper understanding where and when thecitizens of Padang work, live, relax and gath-er. It will develop characteristic scenarios withWP 4200 to create typical boundary conditionsfor the time- and site-dependent numericalevacuation simulations. With these data suc-ceeding simulations of the evacuation behav-

Figure 3: Existing evacuation plan in Padang (Source:KOGAMI)

70

iour within the construction are most likely topinpoint to critical infrastructure (design mal-functions) and bottlenecks within the distinc-tive evacuations procedure. WP 4200 numeri-cally simulates the road traffic in case of anevacuation, using the road network informa-tion provided by WP 3000. The backbone ofthese computations is formed by a so-called»Queue Model«, which simplifies roads asedges and crossings as knots to facilitate quickcomputations of large scenarios. The modelnecessarily demands for simple, graph-basedinformation (connection of the edges andknots; Length, width and capacity of theedges) based on the existing model generatedby WP 2000. Computational evacuation pro-cedures in the streets of Padang also containa behavioural mode including the simulationof individual escape routes. On the whole,future spatial planning conceptions in coastalzones around Padang should make extensiveuse of the outcome of these evacuation mod-els and should connect potential infrastructur-al measures with the hydro numerical one- andtwo-dimensional tsunami inundation model.

WP 5000 provides high resolution spatial data,which is required by WP 2000 and WP 4000.Currently, the only source of surface heightdata in the area is provided through the Shut-tle Radar Topography Mission (SRTM). Themajor disadvantage of SRTM data in contextof damage modelling is the coarse resolutionand non-validated data sources. With thisdata, overview maps produced will not presentprecise modelling of damage patterns. WP5000 focuses on delivering high resolutionaerial images and elevation data of about500 km2 which covers the city of Padang andadjacent villages. Those data are derived byapplying the Multi Functional Camera (MFC)system (developed by the German AerospaceCenter – DLR), which has unique propertiesand is one of the most advanced digital aerialphotographic systems worldwide. Synchro-nous to the image data acquisition with up to16 cm spatial resolution, a detailed Digital Sur-face Model (DSM) is derived through highlyautomated processing. The MFC system is oneof the very few available systems capable ofproducing image and elevation data with theradiometric and geometric quality needed. Thespatial resolution of both, the DSM and theimage data, are 25 cm. Based on the DSM aDigital Elevation Model (DEM) will be derivedand in a further step all buildings with theirindividual heights are being extracted and clas-sified according to their building structure.Those data will be converted by a highly com-puterised process into a 3 dimensional city andcoastal model. Based on this model, thepotential risk on individual buildings can beestimated by taking the hydronumerical calcu-lation (WP 2000) into account. The results willbe implemented into a spatial geodatabaseand combined with results of the WP 1000and WP 3000.In order to evaluate the possible effects oftsunami inundations in Padang in respect toproposed evacuation measures and reductionof potential damages or losses, rule-based(decision-tree) strategies as a function of day-time and weekday information are developed.The consequences of the tsunami flooding atthe coast and in the hinterland in terms of vul-

Figure 4: Last-Mile – Evacuation flight planning to takehigh resolution aerial images of Padang in 2007

71

nerability and hazard assessments are deter-mined by the working packages 1000, 2000and 3000 and made available for all projectstakeholders by a suitable common exchangeinterface. It is of particular interest to deter-mine by which time the streets or individualbuildings in Padang have to be evacuated; WP4100 assesses evacuations within buildingsand typical infrastructure and WP 4200 deter-mines whether the evolving traffic in a height-ened tsunami warning is limited by the pre-sent road network in Padang. In this regard,the present information system can find usagein the distribution of relief efforts after atsunami hit the city, due to fact that street net-work data and flooded districts as well as theessential socio-economic data (demographic,mobility, e.g. availability of cars, motorcyclesor bicycles) are integrated in the system.As an end product, a web based informationsystem will be developed by WP 5000, whichgathers all information derived during the pro-ject »Last-Mile – Evacuation« in order to pro-vide it to public and governmental institutions.The information system will be available in 3Dand displays the potential risks of districts andeven single buildings taking different Tsunami

scenarios into account. It provides all relevantinformation for an evacuation and risk assess-ment in a way which can be easily under-stood, as well as for evaluation of differentmeasures taken in future decision processes,e.g. spatial planning and construction sites,which may decrease the damaging effect ofcoastal hazards. This application will be avail-able via intra- as well as internet. Moreover,video animations are rendered by combiningall components of the project. In this waythe behaviour of a Tsunami which hits thecoast and the wave propagation inside thecity of Padang, as well as potential damageand possible evacuation routes can be realisti-cally simulated. It will be used for awarenesscampaigns and training material for thelocal authorities.

3. Structural Design and Current Statusof the ProjectFigure 6 provide a general impression aboutthe internal structural design of the currentproject. (The detailed list of the German part-ner institutions is attached in the Annex)The five work packages of the project are leadby the respective German partners and each of

Figure 5: Photorealistic 3D city model (Source: RSS GmbH)

72

them is subdivided into several subcompo-nents. Each work package is characterized byinterfaces open for the partner institutions. Bycoupling these components new insights ofthe work packages and subcomponents interms of data exchange or joint scenariodevelopment are created and forwarded tothe overarching work packages Evacuation rec-ommendations & information system and Webapplication that finally lead to the establish-ment of an open software application encom-passing all available scientific knowledge,information, scenarios, data and analysisresults regarding inundation dynamics, vulner-ability and evacuation in Padang. This endproduct is nourished and promoted by allinvolved project partners. Within this projectnetworking is carried out internally as well asexternally. By means of contents-based, tech-nical collaboration of this multi-disciplinarycombination of partner institutions network-ing is done internally.External networking focuses on collaboratingwith the Indonesian partners in GITEWS(BAKOSURTANAL, BPPT and LIPI), yet it prelim-

inary takes the local authorities, the Faculty ofEngineering at Andalas University and theIndonesian research institutions into account.In order to enable a functioning and reliablenetwork of partnering academic institutions,besides signing letters of intent or/and endors-ing the objectives of the project, financial sup-port is also allocated for conducting researchon data collection and analysis of essentialsocio-economic data-sets in Padang. Duringthe foreseen meetings a thorough needs andassessment study of tentative partners inPadang concerning the tangible potentials andexpertise of academic institutions and localauthorities is conducted. Likewise an institu-tional mapping approach on local and region-al scale will be followed to determine mandateand participation in any decision-making pro-cesses regarding disaster management.The above mentioned Indonesian researchinstitutions are already determined throughthe research being executed in GITEWS andled so far to the creation of a well-functioningbilateral network among German and Indone-sian partners. The methods and technologies

Figure 6: Project organization and work structure

73

within »Last-Mile – Evacuation« partially sup-port the work of GITEWS, but take the work-load of GITEWS even further and are not infocus of substantial scientific challenge.This research project is currently in its earlystage, where networking with the localIndonesian institutions is initiated and prelimi-nary data is being collected. A hydrographicalsurvey is currently being conducted in order togather highly-resolved data of near shorebathymetry of Padang to an extent of up to100 km2 as well as collecting informationabout flow rates rectangular and parallel tothe coastline aiming to enhance the insight intidal dynamics and wave interactions. More-over, an official »Kick-off Meeting« with allGerman and Indonesian partners is foreseenby the beginning of 2008.

1 UNDP (2004), Reducing Disaster Risk – A Challenge for

Development. United Nations Development Programme2 MunichRe (2005), Schadenspiegel 3/2005 – Themenheft

Risikofaktor Wasser, MunichRe3 Fernando, J. (2005), Coral Poaching Worsens Tsunami

Destruction in Sri Lanka. Eos, Vol. 86, No. 334, 5 Oumeraci, H. (2003), Wasser im Küstenraum. In:

Denkschrift Wasser, Deutsche Forschungsgemeinschaft

(DFG)6 Bogardi, J. (2004), Hazards, risks and vulnerabilities in a

changing environment. Global Environm. Change 14,

pp. 361–3657 Shah, H. (2003), Last-mile – Earthquake Risk Mitigation

Assistance in Developing Countries, Report Stanford

University8 Population number in year 2005 was 801,344 from BPS

Kota Padang, (2006), Padang dalam Angka 20069 GITEWS Konsortialpartner (2005), Technical Layout of

GITEWS, Helmholtz-Gemeinschaft

Annex I:[1] UNDP (2004), Reducing Disaster Risk – AChallenge for Development. United NationsDevelopment Programme

[2] MunichRe (2005), Schadenspiegel 3/2005– Themenheft Risikofaktor Wasser, MunichRe

[3] Fernando, J. (2005), Coral Poaching Wors-ens Tsunami Destruction in Sri Lanka. Eos, Vol.86, No. 33

[4], [5] Oumeraci, H. (2003), Wasser imKüstenraum. In: Denkschrift Wasser, DeutscheForschungsgemeinschaft (DFG)

[6] Bogardi, J. (2004) Hazards, risks and vul-nerabilities in a changing environment. GlobalEnvironm. Change 14, pp. 361–365

[7] Shah, H. (2003), Last-mile – EarthquakeRisk Mitigation Assistance in DevelopingCountries, Report Stanford University 8

[8] Population number in year 2005 was801,344 from BPS Kota Padang, (2006),Padang dalam Angka 2006

[9] GITEWS Konsortialpartner (2005), Techni-cal Layout of GITEWS, Helmholtz-Gemein-schaft

Annex II: List of German project partners(& institutions) in Last-mile – EvacuationPartner 1: Dr.-Ing. Jörn BirkmannInstitute for Environment and Human Securi-ty, United Nations UniversityHermann-Ehlers-Straße 10, 53113 BonnTel.: +49 (0)228 815–0208,Fax: +49(0)228 815–0299Email: [email protected]

Partner 2: Univ.-Prof. Dr.-Ing. habil. TorstenSchlurmann (Projektleiter)Franzius-Institut für Wasserbau und Küsten-ingenieurwesenFakultät für Bauingenieurwesen undGeodäsie, Leibniz Universität HannoverNienburger Straße 4, 30167 HannoverTel.: +49 (0)511 762–2572,Fax: +49 (0)511 762–4002Email: [email protected]

74

Partner 3: Prof. Dr. S. Dech, Dr. Günter StrunzUniversität Würzburg, Sanderring 2, 97070WürzburgLehrstuhl für FernerkundungEmail: [email protected]

Partner 4a: Dr. Hubert Klüpfel (KMU)TraffGo HT GmbH, Bismarckstr. 142,47057 DuisburgTel.: +49 (0)203 8783–3601,Fax: +49 (0)203 8783–3609Email: [email protected]

Partner 4b: Univ.-Prof. Dr. Kai NagelVerkehrssystemplanung undVerkehrstelematik, Institut für Land- undSeeverkehrFakultät V – Verkehr- und Maschinensysteme,Technische Universität Berlin, Sekr. SG 12,Salzufer 17–19, 10587 BerlinTel.: +49 (0)30 314–23308,Fax.: +49 (0)30 314–26269Email: [email protected]

Partner 5a: Prof. Dr. Florian Siegert,Dr. Claudius MottRemote Sensing Solutions GmbH (RSS),Wörthstr. 49, 81667 MünchenTel.: +49-(0)89–48954765,Fax: +49-(0)89–48954767Email: [email protected] und [email protected]

Partner 5b:Deutsches Zentrum für Luft- undRaumfahrt e.V. (DLR),Linder Höhe51147 Köln(ausführendes Institut: Institut für Robotikund Mechatronik,Abt. Optische InformationssystemeRutherfordstr. 2, Berlin 12489Prof. Dr. Gerd Hirzinger und Frank LehmannTel.: +49 (0)8153 282401,Fax: +49 (0)8153 288667Email: [email protected] [email protected])

Partner 5c: Prof. Dr. Reinhard KleinInstitut für Informatik II – Computer Graphik,Universität BonnRömerstr. 164, 53117 Bonn, GermanyTel.: +49 (0)228 734201,Fax: +49 (0)228 734212Email: [email protected]

75

Sensor based Landslide Early Warning System– SLEWSDevelopment of a geoservice infrastructure asbasis for early warning systems for landslidesby integration of real-time sensors

AbstractEarly warning systems are becoming one ofthe main pillars of disaster prevention in nat-ural hazards especially where mitigation strate-gies are not realizable. Therefore, the call formulti-hazard early warning systems andimprovement of monitoring of natural hazardsis steadily growing. However, data gathering(generation) through sensor integration as wellas system independent supply of informationfrom early warning networks are of particularchallenge.Because very complex informationand warning chains have to be served for,standardisation and interoperability plays animportant role for all organisations and infra-structures being involved. The SLEWS projectinvestigates the complete information chain,starting from data gathering using wirelesssensor networks via information processingand analysis to information retrieval. This isdemonstrated for landslides and mass move-ments. The proposed approach adds especial-ly to the funding target of mobile, cost-reduced and easy deployable measurementsystems, as well as the modern information

systems under consideration of interoperabili-ty and service orientated architecture con-cepts. The proposed wireless network providesthe basis for applying further techniques likesensor fusion, identification of malfunctions orerrors. The obtained geodata is processedaccording to the requirements of particularusers and can be provided for use in a local orregional as well as global information struc-ture. Interfaces will be created, which make amobile and flexible adaptation or integrationof geodata possible. Thus, it is ensured thatthe early warning system reaches the spec-trum of potential users and provides a maxi-mum flexibility.

1. IntroductionDue to the progressive development of urbanareas and infrastructure in Europe as well asworld-wide, more and more people settle inenvironments that are or become endan-gered by mass movements. This situation isbeing complicated by the fact that thedependency of our today’s society on a func-tioning infrastructure and number of human

Arnhardt C. (1), Asch K. (3), Azzam R. (1), Bill, R. (2), Fernandez-Steeger T. M. (1), Homfeld S. D. (4),

Kallash A. (1), Niemeyer F. (2), Ritter H. (4), Toloczyki M. (3), Walter K. (2)

(1) Chair of Engineering Geology and Hydrogeology (LIH), RWTH Aachen University, Email: arnhardt@ / azzam@ /

fernandez-steeger@ / [email protected]

(2) Chair of Geodesy and Geoinformatics (GGR), Rostock University, Email: ralf.bill@ / frank.niemeyer@ /

[email protected]

(3) Federal Institute for Geosciences and Natural Resources (BGR), Hannover, Email: kristine.asch@ /

[email protected]

(4) ScatterWeb GmbH (SWB), Berlin, Email: homfeld@ / [email protected]

76

or objects in endangered areas increases atthe same time. This leads to an overallincrease of risk for our society.This development is not only a specific prob-lem to Germany or the Alpine region. It is aworld-wide challenge apparent from anincreasing number of national and interna-tional programs during the last years. Besidesthe investigation of hazards, these programsare focused on risk reduction by monitoringand early warning, since here immediate pos-itive effects for protection of human lives andobjects can be expected.Currently existing monitoring systems forearly warning are available in terms of mono-lithic systems. This is a very cost-intensiveway considering installation as well as opera-tional and personal expenses. A very complexemergency plan is usually executed in case ofwarning. This requires a disciplined adher-ence to an information chain. Failure ofsingle elements will significantly disturb theflow of information. The new approach pro-vides the possibility for an involved institutionto get user adapted information in a veryearly state independent from a hierarchicinformation structure.The present joint project aims at a systematicdevelopment of a prototyping alarm- andearly warning system to address differentkinds of natural hazards citing landslides as anexample. It is based in innovative service ori-entated information structure integratingOGC standards and the application of highlyflexible sensors, also from automotive areasdue to their readiness and cost-efficiency. It isan advancement regarding classical earlywarning as the sensor networks are free scal-able, extensible and foreign data can be inte-grated from other providers due to an openplatform strategy with web processing ser-vices. The planned geoservice infrastructurewill involve sensors, geodata, recent informa-tion and communication, as well as methodsand models to estimate parameters with rele-vance to stability of landslides.

2. ObjectivesOne main goal of this project is the develop-ment of a low-cost autonmous sensor net-work that is suitable for the detection and theobservation of mass movements. In order toget information about movement rate, accel-eration and movement direction of a landslide-area, appropriate real-time observation sys-tems are selected and will be tested. Onefocus will be on the determination of realisticscenarios and adequate sensor profile require-ments. To develop a reasonable and effectivesensor combination (sensor fusion) and align-ment for early warning, the results of experi-ments with the different sensors will be ana-lyzed and evaluated regarding its quality state-ment. Special focus will be on sensor and net-work fusion with respect to the reduction offalse alarm rates, malfunktion detection andinformation enhancement. The sensors will beintegrated on so called motes, providing ener-gy supply, processing capabilities and a wire-less network connection.In this context, wireless sensor networks makeperfectly sense as they provide a good cover-age of a wide area without inacceptabledeployment effort. With the multihopapproach of wireless networks, each self-orga-nized sensor (node) of the network automati-cally integrates into the network and forwardsboth local measurements and data of peernodes to a data collection.In addition, innovative geodetic measuringprinciples are used to describe movements onthe surface in space and time. The measure-ments resulting in heterogeneous observationssets (real-time and geodetic measuring) haveto be integrated in a common adjustment andfiltering approach. In the given project, geo-detic methods for deformation analyses haveto be adapted and modified with respect toreal-time requirements.The second main goal of this project is to link,prepare and aggregate relevant informationfor early warning systems. This will be basedon data received from different informationsources of the cooperative project partners inthe process of building a Spatial Data and Spa-tial Service Infrastructure. The whole informa-

77

tion System will be optimized on the basis ofthe actual state of concept of web-based, dis-tributed spatial information systems.The development and setup of a geodata andgeoservice infrastructure for the integration ofreal-time data and geoprocessing is essentialfor reaching the objectives. The infrastructuredeveloped here will obtain syntactic andsemantic interoperability. In order to do this,the development of applicable processes willbe required for ad-hoc networking of SpatialServices, and for ad-hoc building of highlycomplex processing routines and models. Theinformation generation will be carried out inparallel to the provided models and processes,but they should remain open for future devel-opment. Much importance will be attached tothe use and further development of existingstandards by the Open Geospatial Consortium(OGC), whereby synergies with other projectscan result.Another important goal of the project is thefformulation and analysis of an interface toend users and to political decision makers. Anessential component is the integration of thesystem into present network structures, the

connection to other early warning systems andthe involvement of attained results into na-tional and international research programs.Working packages were defined and eachpackage is assigned to a project partner, whois responsible for content processing. TheDepartment of Engineering Geology andHydrogeology of the RWTH Aachen University(LIH) is responsible for establishing an obser-vation network integrating and evaluatingnew geo-sensors for the observation of massmovements in consideration of landslide typeand mechanism. he ScatterWeb GmbH (SWB)manages the initialization of wireless, self-gov-erned and self-organizing sensors with com-petitive costs. The Department of Geodesyand Geoinformatics of the Rostock University(GGR) improves data quality and provides theinterface for implementing a geoservice infra-structure (GDI), while the Federal Institutionfor Geosciences and Natural Resources (BGR)identifies users demand and the possibility toprofit by the project’s outcome by installing anexpert system comparative to European stand-ards. Although some partners leading tasks inthe frame of the joint project the partners will

Figure 1: Linkage and interaction between project partners involved

78

work in close cooperation to ensure a smoothflow of information and exchange to ensureprogress (Fig. 1).

3. Technology

3.1. Wireless sensor network

3.1.1. Sensor concepts and sensor fusionA new level of miniaturization of processorsand radio modules encouraged the rise ofa new kind of networks: ad hoc wireless sen-sor networks. Measurement data is collectedby distributed sensor nodes, which interactseparately and connect each other in a self-organizing manner. Modern nodes for wire-less sensor networks are built in a stronglymodular way by providing open interfacesfor integration of very different measuringsensors. Simple temperature and humiditysensors can be integrated as well as high-pre-cision displacement and acceleration trans-ducers, tiltmeters , geophysical and acousticsensors or even GPS modules for locationdetermination (fig. 2).For the observation of mass movements it isvery important to define which measuring sen-

sor could be used for a special failure mecha-nism. In order to observe slow movements,like creeping, very sensitive measuring sensorsare necessary, that can detect small elonga-tions. These sensors are often very expensive,because of their high precision. By contrastslide and fall movements are much faster, witha higher acceleration and elongation. Hence,sensors are used that can detect larger move-ment amplitudes. This knowledge is associat-ed with selection of the sensors and the tar-get of this project in respect of low cost sen-sors. The idea is to use sensors which are easyto get from the automotive industry with lesscomplexity in integrating them to sense themass movements.Anyway in the beginning it is important to linkbetween the sensor measurement and the fail-ure mechanism. Tab. 1 shows the five mainlandslide failure mechanisms that are basicallyexamined in this project. The reasons to turnthe attention to these failure mechanisms are:1. The project concentrates on fast mass

movements like rock fall and topples orrock slide.

2. Debris flow will not to behold because ofthe coasts of the monitoring measurements

Figure 2: Schema of a Sensor based Landslide Early Warning System

79

and completely different boundary condi-tions (flow process, very fast).

3. The warning should be a short term or postevent warning.

As stated in Tab. 1, the sensing of the accel-eration is helpful in all chosen mechanises Dueto the kind of the used sensor (three axial ormono axial) the signal could be dashed or con-tinues in the case of fall or topple. If a threeaxial sensor is used the signal will be continu-ous but it will be dashed if the sensor has onlyone axis (Fig. 3). Because of the high price ofthe three axial sensors, three or two mono-axial sensor will be orientated monted on a

mote to set up a multi axis accelaeration sen-sor.The sensing of displacement can be realisedusing a potentiometric Draw Wire Displace-ment Transducers or with normal Linear Dis-placement Transducers (Fig. 4).The Draw Wire Displacement Transducer hasthe advantages that the stainless steel cable isrelatively flexible at its end, so that a move-ment in all directions will be no problem.The Angle sensor is a useful instrument forthe monitoring of toppling or in some casesfalling. In the case of rotational slide if theslip surface is known, then it will be useful

Table 1: Example of Failure Mechanisms and Instrumentation of different Measuring Sensors

Figure 3: Examples of diagrams from mono- axial and three-axial acceleration sensors

80

to use an angle transducer to identify theactive part of the slide. Angle sensor canalso help to detect reactivation of endan-gered ares as a first signal for beginningmovments. Additionally angle sensors can beused to measure landslide movements indi-rectly on walls or buildings for example, asthey detect the deformation and tilting ofsuch constructions.In the frame of the project, selection and test-ing of appropriate sensor configurations andfusion of data from different sensors (usingsensor fusion) to improve prediction quality isof major importance. The focus lies on thedetermination of realistic scenarios and ade-quate sensor profile requirements.The results of realistic experiments with dif-ferent sensors will be analyzed and evaluat-ed regarding its quality statement. Mainaspects are:– Significance of sensor signals with regard to

geology and mechanics– Improved gathering of information through

sensor and network fusion (fusion of diffe-rent data)

– Identification and handling of errors, outliersand malfunctions

– Evaluation of measurements, data fusionand realization of decision making conceptswith regard to false alarm rate reduction

– Examination and evaluation of the sensornetwork concerning applicability as monito-ring tool beyond early warning

In cooperation with the project partners sim-ple but effective decision making algorithmswill be developed to evaluate the results fromexperiments. Thus, early warning may be real-ized under outdoor conditions. For the algo-rithm development hierarchic decision makingstructures (e.g. simple exclusion rules like»winner takes it all« or majority decision,regression trees) will be used due to theirinterpretability and feasibility. Alternatively,cluster- and fuzzy logic or logistic regressionmethods will be examined. Data mining meth-ods will be used for data analysis and param-eterization of decision making processes.

3.1.2. Network and nodesModern measurement nodes for wireless sen-sor networks are built in a strongly modularway. That way they especially provide openinterfaces for integration of very different sen-sors. Simple temperature and humidity sensorscan be integrated as well as high-precisionvibration sensors or even GPS modules forexact location determination.Currently available products are mainly exper-imental platforms with only a few of themready for the market. Products that are readyfor the market have to pass certifications ofregulatory bodies. Especially the Europeanstandard EN 300-220 has to be obeyed forany low-rate radio module that may be usedoutside research laboratories. Certificationsalso include further market relevant assertions

A.) Draw Wire Displacement Transducer B.) Linear Displacement Transducer

Figure 4: Examples of displacement transducers

81

Figure 5: ScatterGate, measurement nodes and case in aruggedized version

Figure 6: SensorGIS installation in the Swiss alps

like reliability even in a wide temperaturerange from –20 °C to 75 °C. The ScatterNodesof the project partner ScatterWeb follow theserequirements and are ready to be used.Sensor data is then collected by distributedmeasurement nodes and forwarded across thewireless network towards a data collectionpoint. This collection point aggregates dataand may perform further steps like data pre-processing and compression. Finally, data is

provided via a wide-area-network likeGSM/UMTS for secured remote access via theInternet. In the ScatterWeb product line, thedata collection point is called the ScatterGate.A set of measurement nodes and one Scatter-Gate in a ruggedized version can be seen inthe following figure.Low energy consumption is a key requirementfor long-living wireless sensor networks.Therefore, modern sensor networks usepower-down modes of processor and radiomodule as much as possible while still provid-ing fast and reliable data transmission. Thatway it is possible to run an outdoor wirelesssensor network for years using standard bat-teries or solar cells, where possible. The fol-lowing picture shows the installation of a WSNusing components from ScatterWeb for thestudy of warming effects in the Swiss alps aspart of the project SensorGIS [15].

3.1.3. Wireless sensor network technology inlandslide monitoringMass movements, e.g. slope movements orrock fall, may be described as complex defor-mation processes at the surface, resulting in

82

a down-slope movement. The observationand monitoring of such movements needs alarge amount of observations (in-situ meas-ured data) and information. Only the com-bination of such information permits anappropriate conclusion about mechanism andattitude of the sliding area resulting in earlywarning information.The Wireless Sensor Network (WSN) consistsof inexpensive computational nodes whichgather relevant parameters and transfer thisinformation to a collection point for furtherprocessing. WSNs nodes are used for sensingthe environment while at the same time theycommunicate with neighbouring nodes andcan even perform basic computations on thecollected data. The usage of WSNs coversover than 222 of general and special appli-cations [1]. The usage of WSNs in engineer-ing geology is at the moment at the begin-ning but the following examples show theworldwide interest in the application on engi-neering geological and other problems inhazard monitoring.One of these projects is »SENSLIDE« –Projectin India leaded by US universities. »SENSLIDEis a distributed sensor system for predictinglandslide« [2]. The basic idea of this project isto use a large number of distributed inexpen-sive single-axis strain gauges connected to acheap node. The missing linkage betweensensors and the punctual measuring methodof these sensors inform about rock movementin various parts pointly only, without informa-tion about relative motion between therocks.Thus, »by measuring the cause of thelandslide, we can predict landslides as easilyas if we were measuring the incipient relativemovement of rocks« [2].Another project uses WSNs to predict the loca-tion of the slip surface of the landslide. Toinvolve the sensors this project build a »SensorColumn«, which includes four types of sen-sors: geophones, strain gages, pore pressuretransducer and reflectometers. The WSNMotes have the function of a telemetric sys-tem to transmit informations from the singlecolumn into the network.

The project »Wireless Sensor Networks withSelf-Organization Capabilities for Critical andEmergency Applications (WINSOC)« aims inone of its sub projects to use WSN to detectrainfall induced landslides and lahars in India.This Project is relatively a large one with 11partners in 7 countries [4]. The project aimsthe development of WSN in early warning sys-tems including hardware and software devel-opment under consideration of biologicalinspired information processing. The mainidea here is to use large amounts of inexpen-sive sensor nodes which collect data and com-pute them in a reasonable local decision mak-ing processes [4].The above described projects are all at the firststages of realisation. Different concepts ofstrategies are under consideration to applyWSN to hazard observation and early warningsystems, from single use as telemetric systemto integrated information processing net-works. This shows that the application is atthe moment still in the beginning but has agreat potential, especially as the technologyWSN seems to become level of good maturi-ty. Especially the evolution of Micro sensorse.g. from automotive sector and the applica-tion of sensor fusion offer a great potential.

3.2. Information system

3.2.1. State of the artStudies and pilot applications, in particularwithin the Open Geospatial Consortium(OGC), have shown that the classic technolo-gies (measurement systems, simulation/calcu-lation models, database systems, mappingand geographic information applications)should be extended or replaced by interrelat-ed infrastructure-oriented technologies whichmay be described using the keywords OpenWeb Services, Sensor Web and Web Process-ing Service (WPS). These are grounded in aweb-service paradigm and have been devel-oped in the context of OGC Spatial Webapplications. This service-orientated technolo-gy is overall highly interoperable and morecapable of being coupled together than theclassic monolithic technologies.

83

Service implementations are made available onthe internet as cascading and self-describingprocesses. This means that a service may becalled not only directly by a human user, butmay also be called from another service. Eachservice may describe what operations are exe-cutable, what inputs are required and whatoutputs are possible in both human- andmachine-readable form. Whole productionchains for the preparation of information maybe realised using these chained or cascadingservices. Because each service is capable ofdescribing its own operations, the cascademay be formed not only as a linear structurebut also as a complex and partially self-organ-ising network. These services therefore alreadyimplement structural elements of the semanticweb, the successor of the WWW.Web services isolate single tasks for spatialinformation management and implementthese as separate modules within a singlearchitecture. These modules take the form ofself-contained services which are capable ofintercommunication via standardised webinterfaces. Such OGC Web Services (OWS)form the basis of spatial data infrastructure(SDI) projects on the international (GSDI) andnational (GDI-DE) level, as well as in almost allfederal states (e.g. GDI-NRW, GDI GeoMV).Currently the most widely-implemented andsophisticated services are the Web Map Ser-vice (WMS), Web Feature Service (WFS) andWeb Coverage Service (WCS). Further servicesare e.g. the Web Catalog Service (CSW) andthe Web Coordinate Transformation Service(WCTS). OGC specifications for Sensor WebEnablement (SWE) and WPS are also available,some currently in draft form, others alreadyapproved. These broaden the OWS to includesensors and models as well as processing andanalysis functionality. Initial pilot studies, suchas the OGC Web Services Testbed Phases 3and 4, have indicated that the developmentof the concept has reached an application-ready stage.

3.2.2. Information system architectureThe prototype of a Spatial Data Infrastructure(SDI) in the context of an alarm- and early

warning system will be orchestrated by anumber of services described by the OGC.Applied will be services of the OGC SWE fam-ily such as Sensor Observation Service (SOS)and Sensor Alert Service (SAS) accompaniedby the use of associated specifications suchas Sensor Model Language (SensorML) andObservations & Measurements Schema (O&M)and other services for the use of mappingdata (WMS), meta data (CSW) and vector andraster data (WFS, WCS). Another fundamen-tal task will be the implementation of analyt-ic processes and calculations forming thebusiness logic represented in self-containedweb-based services (WPS).Early drafts encapsulate the information infra-structure in three main parts: sensor-sidelogic, server-side logic and client-side logic.The sensor-side logic contains process chainsto survey and accumulate sensor data repre-senting the underlying sensor network to theoutside via SOS et al. at a level of abstractionstill to be defined. The server-side logicembodies the infrastructure core by providingprocess chains for harvesting, analysing andvisualizing sensor data and sensor parametersunder consideration of geodetic and engi-neering geologic expertise. The client-sidelogic will provide a combination of portal andcatalogue services allowing decision-makersto research and visualise security-relevant datafrom landslide-endangered areas. Visualisa-tion techniques will be designed to take theuser’s knowledge related to the field intoaccount and will range from detailed outputto fuzzy graphical warning elements. Addi-tional synchronous and asynchronous infor-mation services as SAS and the Web Notifica-tion Services (WNS) can provide defined usersand user groups with time-critical readingsfrom the observation site.Despite a certain predetermined process chainbetween measuring sensor data and formulat-ing conclusions for information purposes, dueto its open and standardised architecture thereare no constraints in using the described ser-vices in an arbitrary manner. Enabled throughstandardised software service interfaces, expe-rienced users can employ the provided services

84

to fit their needs using a wide range of exist-ing compliant software products.

3.2.3. Positioning and localizationOne aim of the project »SLEWS« is to devel-op low-cost-methods and techniques forearly warning systems against landslides.There are possibilities to use special sensorssuch as inclinometer, pressure sensor, hygro-meter, humidity sensors or linear positionsensors. These sensors have high accuracy,for example less than one millimetre. Priceand accuracy stand in direct relation. Allthese sensors are local sensors with outputabout local phenomena.Classical surveying methods such as GlobalPositioning System, terrestrial measurementand remote sensing are established means toget information about movements across thearea of unstable land. Although such methodshave a high accuracy, their application is veryexpensive and they will therefore only be usedin areas where the potential costs of a land-slide are much higher. All these methods needa direct view between sensors and reflectorsand/or receivers. The direct visible connectionwill be influenced for example by vegetationand is not always assured. Furthermore, thetotal cost of the measurements increases witha decreasing interval between epochs.Many observation points must be defined tocover the area. The outcome of this is a localgeodetic network. Geodetic networks can bemeasured with the help of electro-opticalrange finding and theodolites. The accuracies

of electro-optical range finding today are atthe millimetre level. The accuracies for thebest types of theodolites today are under onemilligon (3 seconds). The reference points ofthis local area network will be realised withdifferential GPS. Accuracies of a few millime-tres are possible.It is also possible to measure distances withother methods such as radio technology. Oneof the newest methods in this domain isWLAN (WLAN Indoor Positioning System).Tests of WLAN Indoor Positioning Systemshow accuracies of positioning between 1 and3 meters. For this reason, this method isunsuitable for monitoring land slips. To getacceptable information about land slips, themeasuring methods must be accurate to onecentimetre. Therefore, ultra-wide band (UWB)technologies, which promise higher accuracies(sub-centimetre), may be applicable. The fol-lowing paragraphs detail some importantresults of studies concerning UWB.Radio systems use electromagnetic waves withdifferent frequencies to send information. Oneof these systems is the narrow-band system,which is also used in electro-optical rangefinding for surveying. The distance is calculat-ed using the difference of phases on sensorand receiver. The accuracy in this case is lessthan one millimetre. To use this system it isnecessary to have a direct view between sen-sor and reflector. Multipath effects are a com-mon cause of error in this method.The next system is the broad-band system,also be known as »Multi Band Orthogonal Fre-

Figure 7: Information system service infrastructure:Sensor-side logic: (a) business logic (b) SOS (c) SAS/WNS;Server-side logic: (d) business logic/WPS (e) data base system (f) WMS/WFS/WCS;Client-side logic: (g) visualization services (h) CSW (i) viewing/portal services and applications

85

quency Division Multiplexing«. In this casethe information sequence is modulated on acarrier frequency. The receiver demodulatesthe information from the carrier frequency.This system is used for example in GPS (C/ACode). Multipath effects are also a problemfor this method.Ultra-wide band has the possibility to sendinformation with or without a carrier frequen-cy. The absolute bandwidth is more than500 MHz. The relative bandwidth is a quarterof the middle frequency. The frequency rangesbetween 0,1–0,96 GHz and 3,1–10,6 GHz butthe information is sent with a low-power(max. –41,3 dBm/MHz). The maximum rangefor this method is therefore 50 metres. Tosend information without a carrier frequency,the information must be discretely encoded asbinary data. Due to the high frequency andvery small electro-magnetic pulses, a high datarate (theoretical more than 1 GBit/sec) can beachieved. This standard was developed in twoworking groups at the Institute of Electrical anElectronics Engineers (IEEE). The first group802.15.3 developed standards for pulsemethod (Direct Sequence UWB, DS-UWB) andthe second group 802.15.4 developed stan-dards for the system with carrier frequency(Multi Band Orthogonal Frequency DivisionMultiplexing, MB-OFDM). In 2002 the FederalCommunications Commission (FCC) releasedUWB licences for free use. Two Europeanworking groups at the European Telecommu-nications Standards Institute (ETSI) have devel-oped standards for the European market(ERM/TG31A, ERM/TG31B).BLANKENBACH et al. [6] carried out tests withUWB for distance measurement. The teststook place outside with direct view. The resultshave been compared with a known true dis-tance measurement. A number of distanceswere each measured 100 times. It was possi-ble to measure distances of up to 50 meters.BLANKENBACH et al. gives 7cm for the standarddeviation of one observation and 1 centimetrefor the standard deviation of the arithmeticmean. Differences between the true andobserved values due to systematic errors wereless than 10 centimetres.

The most important advantages and disadvan-tages are as follows:

Advantages– very small electro-magnetic pulses (<2 nano-

sec) no multi path effects– signals saturate materials– UWB signals are mostly not distinguishable

from background noise for other radio recei-ver because of restriction of transceiverpower

– UWB systems do not need a carrier frequen-cy

– The installation and the hardware are easierand cheaper compared to systems with acarrier frequency

– less power requirement compared to othersystems with a carrier frequency (for exam-ple WLAN)

– no direct view is needed

Disadvantages– maximum range is about 50 m.– it is possible, that UWB transmitters disrupt

other frequenciesSome questions remain currently unan-swered, such as:

– Are there instruments for commercial use?– How much will they cost and what is their

accuracy?– Who sells such systems?UWB is a new technology, which can be usedto transmit data with high rates. It is also pos-sible to use UWB as a range-finding method.The accuracy is high enough to consider theuse in landslide monitoring.

3.2.4. Work progressRecent progression has been made in design-ing a generic extensible data model for gath-ering positioning data. At this point the datamodel contains geodetic information neces-sary for determining angles und distancesbetween observation points, informationwhich is used to express possible surfacedeformation processes. Future data sets canbe fitted with an arbitrary amount of addi-tional information attributes. A data profilehas been implemented in the spatial database

86

management system PostGIS and data setscan be visualised using WMS provided byUMN MapServer. The data profile has been fit-ted with real-life sample data to plot firstapplication scenarios.Future steps will address first deployment ofOGC SWE services, for example describingpreliminary sensor setups in SensorML andmaking remote simulated sample data avail-able via SOS.

3.3. Warning and end userEarly warning systems require very complexinformation and communication structures.They need to interface sensor technology,data management with data base manage-ment systems, data preparation and analysis,prognosis and model calculations, computernetworks and communication terminal units,and also result in a processing and communi-cation tool.Studies and pilot trials, particularly within thecontext of the Open Geospatial Consortium(OGC), have shown that classic technologiesare replaced by interrelated, infrastructure-focused technologies. They are all based onthe Web Service Paradigm and were devel-oped within the context of the OGC’s SpatialWeb approach. This technology has far greaternetworking and interfacing capabilities thanthe classic technologies. Initial pilot tests indi-cate that the concept has reached a level of apossible real world implementation. It has tobe evaluated how far developed concepts canbe implemented in practice and which modifi-cations and changes are required in order touse the new technology successfully for earlywarning purposes.

In this joint project the Federal Institute forGeosciences and Natural Resources (BGR) Han-nover is embedded as a multiplier to the com-munity as well as a potential user of an earlywarning system for landslide hazards. Theinvolvement of potential users in this earlystage of the project ensures the developmentof a high quality, re-usable and portable infor-mation system for application in the field ofearly warning.Potential users may be dam operators whomonitor unstable watersides (e.g. ThüringerTalsperrenverwaltung, slope Gabel) or the Ger-man Rail Company (DB) which needs to main-tain the stability and safety of its rail network(e.g. rock fall hazards on Rheintal route).Other potential users are operators of moun-tain railways and lifts in the Alps region.Repeated thawing of permafrost can cause de-stabilization of subsoil. A second relevantgroup of users are those responsible for iso-lated sensor networks or early warning sys-tems whose applications might integrate datainto the system. Examples for these are damoperators, mining companies (RWE Power),the BGR and even participants of the BMBFprogram (GFZ, AWI). Besides these potentialusers operators of early warning systems forother natural hazards in regard to mass move-ment will have an interest in this study. Exam-ples of these are the measuring network andearly warning systems of the German Weath-er Service or the flood water central Baden-Württemberg, which operates a measuringnetwork for rainfall and river gauges related toflood forecasts.The project results are planned to comply withopen international specifications to ensure

Figure 8: Digital elevation model sample data (left), PostGIS profile (center), visualization via WMS (right)

87

their availability to a wide community. As theWPS specification of the OGC is adopted itcould advance to an international standardwithin the scope of this project. The projectgoals are beside others the determination ofneeds and the development and configurationof end-user friendly early warning-systems.Finally the integration of the project-resultsinto national, European and international pro-grams should be realized following interna-tional standards.An important aspect of the development ofinformation systems is to observe standards inorder to be able to obtain transferability onother questions and reuse of developed tech-nologies in future applications. The resultshave to be integrated directly into open andinternational specifications in order to makethem available to a broad Community. Thedevelopment has to be increased to establishdeterministic prediction systems supplyingearly warning systems with reliable data con-sidering only a few determinable parameters.This is of importance, because monitoring sys-tems tend to thin out with the times due tothe loss of hardware components.

4. ConclusionThis project provides a substantial contributionto the development of methods and tech-nologies for early warnings with regard tomass movement events. Its interdisciplinaryconstitution offers the possibility to processthe complex problem of effective early warn-ing by classifying multiple approaches. Highlycomplex events like slopes failures depend onmultiple sensitive parameters (e.g. humidityand stress constitution in an endangered area).The combination of small and precise measur-ing sensors like tiltmeter or displacementtransducers, used in automobile technology,with a self-organizing monitoring system in anetwork permits the advance of a real-timemonitoring system that is suitable for thedetection and the observation of mass move-ments. Modern sensor networks, like ad hocwireless sensor networks have the advantageover currently existing landslide monitoringsystems that they can be used very variable

and the installation is quite simple. Further-more the short-distance communicationbetween adjacent nodes (still up to 1 km freespace communication distance) allows cuttingdown costs and administrative effort. In addi-tion, as the nodes are directly connected witheach other, local preprocessing of the data canbe applied and sensor fusion techniques canminimize false positives.The planned geoservice infrastructure willinvolve sensors, geodata, recent informationand communication, as well as methods andmodels to estimate parameters with rele-vance to stability of landslides. The process-ing of these geodata applying capable proce-dures and algorithms will lead to better pre-dictions and early warnings. By these aspects,this project contributes important improve-ment of geoservices by developing geoinfor-mational methods. The new approach pro-vides the possibility for any involved institu-tion to get user adapted information in a veryearly state independent from a hierarchicalinformation structure. The service orientatedinfrastructure closes the gap between datagathering and information retrieval using ser-vices as SAS and WNS. The intermediateresults of the project are linked to the feed-back of potential users, which will be inte-grated in the current project via workshopsand interviews. Improvement of user inter-faces and possibility of model and prognosisintegration are of main interest and will leadto the development of new standards.

5. References[1] K. Sohraby, D. Minoli and T. Znati (2007):Wireless Sensor Networks: Technology, Proto-cols, and Applications, by John Wiley &Sons, Inc.

[2] A. Sheth, Ch. A. Thekkath: SenSlide(2005): A Sensor Network Based Landslide Pre-diction System SenSys, 05, Nov 2–4

[3] A. Terzis, A, Anandarajah, K. Morre and I,Wang (2006): Slip Surface Localization inWireless Sensor Networks for Landslide Predic-tion, IPSN’06

88

[4] http://www.winsoc.org/keyfact.htm

[5] Retscher, Günther; Moser, Eva: Genauig-keits- und Leistungstest eines WLAN IndoorPositionierungssystems. Zeitschriftenartikelaus: zfv Zeitschrift für Geodäsie, Geoinforma-tion und Landmanagement, Jg.: 132, Nr. 1,2007, S. 4–10

[6] Blankenbach, J.; Norrdine, A.; Schlemmer,H.; Willert, V.: Indoor-Positionierung auf Basisvon Ultra Wide Band, Zeitschriftenartikel aus:AVN Allgemeine Vermessungs-Nachrichten,Mai 2007, S. 169–178

[7] IMST GmbH: UWB – Ultra Wide Band. UR:http://www.imst.de/de/funk_wir_uwb.php.Letzter Zugriff am 06.07.2007

[8] Bartels, Oliver: UWB: Störer oder Helfer?URL: http://www.heise.de/mobil/artikel/57048(04.03.2005). Letzter Zugriff am 06.07.2007

[9] ITWissen; UWB (ultra wideband). URL:http://www.itwissen.info/index.php?aoid=13589&id=31. Letzter Zugriff am 06.07.2007

[10] CIO: Trapeze offers 802.11n draft accesspoint. URL: http://www.cio.de/news/cio_worldnews/838319/index1.html (19.06.2007).Letzter Zugriff am 06.07.2007

[11] Macwelt: 480 Mbit/s: Ultra-Breitband-Chip für Mega-Content-Handys. URL:http://www.macwelt.de/news/netz/443866/index.html. Letzter Zugriff am 06.07.2007

[12] ELKO: UWB – Ultra-Wideband Wireless.URL: http://elektronik-kompendium.de/sites/kom/1010131.htm. Letzter Zugriff am06.07.2007

[13] Sikora, Axel : Ratgeber: Funknetzwerkedaheim. URL: http://www.pcwelt.de/start/com-puter/netzwerk/praxis/85162/. Letzter Zugriffam 06.07.2007

[14] Macwelt: ZigBee-Standard verabschie-det: Revolution der Fernsteuerung?. URL:http://www.macwelt.de/news/soft-ware/335987/index.html. Letzter Zugriff am06.07.2007

[15] http://www.sensorgis.de

89

Integrative Landslides Early Warning Systems(ILEWS)

IntroductionEarly warning of landslides is a challengingtopic because the possibilities of predictionvary significantly. A distinction is not only lim-ited to different types of landslides, it alsocomprises the differentiation between reac-tivated and new movements. Due to theiralready moved compound, reactivated land-slides are relatively easy to identify and ifrequired, appropriate monitoring instrumentscan be installed. The locality of new landslidesis by contrast rather difficult to predict.The aim is to design and implement an inte-grative early warning system for known (reac-tivated) and new landslides and debris flows,which provides information on future eventswith regard to local and regional require-ments. The methodical configuration of theearly warning system is developed to be trans-ferable and modular, i.e. it can be adapted tolocal structures of different countries, and itcan be customized to other natural processes(e.g. rockfalls). Key project targets are:– Formulation of an integrative early warning

concept for landslides– Investigation and installation of an adapted

early warning system– Monitoring, parametrisation and modelling

of local data

– Development of risk management options– Linkage of local findings to regional model-

ling– Provision of information necessary for early

warning– Integration of warning into the respective

social processes of decision making.Strengthening the awareness towardsunderestimated risks that are associatedwith landslides

– Transfer of the concept to an area withmany, already existent monitoring stationswhich are not yet networked in the abovementioned manner.

Within the scope of the submitted cooperativeproject, early warning systems are to be imple-mented in two European test areas.

1. Swabian Alb in GermanyThe study area is a settlement area on a his-torically active complex rotational slide. Reg-ularly recurring damage to houses and meas-urements of inclinometer show the reactiva-tion of at least parts of the slide. The land-slide body is already under investigation with-in the InterRISK research project, so that theexisting infrastructure as well as valuabledata can be accessed and continuativeknowledge may be built.

Glade T. (3), Becker R. (6), Bell R. (3), Burghaus S. (2), Danscheid M. (2), Dix A. (1), Greiving S. (7),

Greve K. (2), Jäger S. (5), Kuhlmann H. (2), Krummel H. (4), Paulsen H. (8), Pohl J. (2), Röhrs M. (1)

(1) University of Bamberg

(2) University of Bonn

(3) University of Vienna

(4) geoFact

(5) geomer

(6) IMKO

(7) plan + risk consult

(8) terrestris

90

2. South Tyrol in ItalyResearch focuses on a debris flow in Nals,which is already equipped with an early warn-ing system and the Corvara Landslide, a well-investigated complex rotational slide flowwhich poses a threat to the village of Corvara.The monitoring infrastructure and long seriesof measurement of the debris flow and theCorvara Landslide are provided for the projectfree of charge.In a first step it is intended to implement anideal type of early warning system in the studysite of the Swabian Alb – from sensor to rec-ommendation of actions. In a second step it isattempted to apply findings and methodsfrom this early warning system to the secondstudy site in South Tyrol.The early warning system consists ofthree clusters: Monitoring, Modelling andImplementation.The Cluster Monitoring is designed to realise arobust and easy to handle online-measure-ment-system for comprehensive monitoring oflandslides. It provides the basis for an integra-tive early warning system. The system consistsof a new sensor-combination for the divisionsmeteorology, soil hydraulics and –mechanicswith the corresponding innovative data trans-mission technologies. The broad spectrum ofobserved processes contributes to a betterunderstanding of process interdependencies.Hence, the evaluation of the risk of criticallandslides can be significantly be improved.Measurement and data transmission are auto-mated. Data is provided in a standardised formby a central measurement database for furtherprocessing (CoreSDI, Setup Monitoring, Sen-sorGIS). Important design criteria for the sur-vey system are the ability to operate on differ-ent scales and to be transferable. Using adjust-ed sensor systems for various area sizes,regions, and processes, regional conclusionsmust be able to be drawn.The main aim of the Cluster Modelling is theconversion of all continuously and periodicallygathered information into a reliable and effi-cient early warning. Therefore, different newmodels will be applied (Movement analysisEarly Warning Model, Physically based Early

Warning Model) and an already existing model(Real-Time Early Warning Model for DebrisFlows) will be integrated.The core of the Movement analysis EarlyWarning Model (Geomorphic Modelling, Geo-detic Modelling) is the trend analysis of thereciprocal landslide movement rate. Lineartrends give an early indication of the time fora catastrophic landslide event. This model iscomplemented by the integration of the con-tinuously measured field parameters into per-manent slope stability modelling. The latterresults into a permanently updated safety fac-tor (Physically based Early Warning Model).The combination of both early warning mod-els should lead to a more reliable early warn-ing. Regarding debris flows an existing RealTime Early Warning System will be used andoptimised. The disadvantage of the existingsystem is the short premonition time of onlysome few minutes. It is intended to extendthis time span by the integration of the weath-er forecast. Particularly important is the designof a structure, that is still fully functioning evenif one or more system components fail.The Implementation of an early warning sys-tem is a multistage process. It starts with thedefinition of protection goals – differentiatedfor the various subjects of protection and theirrespective damage potential (economic values,critical infrastructure, people, environment).The defined protection goals are then animportant basis for decision-making of theClusters Monitoring and Modelling.The scientific analysis is an essential part of anintegrated early warning system. However,the question which risks are acceptable or tol-erable is a normative one, i.e. it has to beanswered by the authorised players. Thus, acooperative risk communication (Communica-tion) is integrated into the cooperative pro-ject, identifying the configuration of theinvolved actors and the communication rela-tionships between these. Within the frame-work of an information management, respec-tive results will be the basis for the develop-ment of a communication network towardsand between the involved actors of a earlywarning chain.

91

The described project ILEWS is divided into tensubprojects. Five of them are executed by uni-versities, the other ones by companies. Theaims of the different subprojects will beexplained on the next pages.

Subproject: Monitoring of landslidemovement and Early Warning Modelling

Thomas Glade & Rainer Bell,University of Vienna

Within this subproject shallow translationalslides, deep-seated rotational slides, and debrisflows will be investigated. Subsurface landslidemovements are monitored periodically using amobile inclinometer device and continuouslywith a permanent installed inclinometer chain.Gained results complete the assessment of rel-evant early warning parameters for theprocess type »slide« within the Cluster Moni-toring. All gathered monitoring data of theSwabian Alb and supplied data of South Tyrolare integrated for an user optimised and reli-able early warning message. The early warn-ing modelling concept uses a physically based»Near Real-Time« Early Warning Model, a Sur-face Movement Analysis Early Warning Model(with the subproject »Geodetic Modelling«)and Regional Early Warning Models.Within the physically based »Near Real-Time«Early Warning Model the continuously meas-ured data, especially soil moisture and rainfall(provided by the subprojects »Moisture Geo-lelectric« and »Setup Monitoring«), will beintegrated in equations to calculate slope sta-bility. Thus, a continuous safety factor can becalculated (the respective WebGIS applicationwill be programmed by the subproject »Info-Management«). If the safety factor gets lowerthan a specified threshold value, preliminarywarning messages are provided in theWebGIS and SMS are sent to the scientists ofthe subproject (by the subproject »Sensor-GIS«) to check and validate the warning usingsophisticated slope stability models as well ascurrent subsurface movements. The slope sta-bility software will also be applied to calculatehighly likely sliding circles for rotational slides

and to identify the best suited slope stabilityequations. The results are then again input forthe physically based »Near Real-Time« EarlyWarning Model. At the end of the optimisa-tion period an autarkic running early warningmodel will be set up, which controls to somedegree itself and must only be supervisedby experts.Within the Surface Movement Analysis EarlyWarning Model all measured movementrates (inclinometer, inclinometer chain,tachymetry, GPS – with subproject »GeodeticModelling«) are analysed using the approachof »progressive failures« (Petley et al. 2005).Depending on the way how the movementrates change, the catastrophic failure of aslope can be predicted. In a last step, it isinvestigated if both models can automatical-ly support each other.It is tested at the Reisenschuh landslide inSouth Tyrol, whether the early warning mod-els developed for slides in the Swabian Albcan be transferred to other study areas, whichare already partly equipped with measure-ment instruments. Regarding the alreadyexisting Debris Flow Early Warning System thescientific basis as well as the extensive meas-urement series will be analysed in detail tosuggest possibilities to improve the system onthe one hand. On the other hand importantinsights towards the set up of a Regional EarlyWarning Model (with the subproject »Histo-ry«) are expected.Local Early Warning Modelling will also pro-vide information on the magnitude of poten-tial events. To better define the endangeredareas depending on the type and magnitudeof the process empirical or physically basedrun out models are applied. Regional EarlyWarning Models for slides and debris flows arebased on the regionalised critical local condi-tions. Regarding slides GIS-based models willbe applied which calculate the threat due totranslational slides on the basis of the »infiniteslope model«. For debris flows an statisticalmodel will be applied, which transfers theresults based on the extensive measurementseries from one debris flow track to othercatchments. The integration of the official

92

weather forecast into local and regional earlywarning models will lead to an expansion ofthe warning time. A comparative analysis of allapplied early warning models will give impor-tant information on the usage of eachapproach for different conditions.Historical frequency-magnitude analyses basedon the data provided by the subproject »His-tory« will help to estimate the recurrenceinterval of landslides of various magnitudesmore reliably and to better interpret localmeasurement results. The GIS-based combina-tion of the early warning modelling and dam-age estimation (with the subproject »Manage-ment«) will provide very helpful informationfor decision makers for choosing the best-suited system for risk mitigation and reduc-tion. In cooperation with the Cluster Imple-mentation a user optimised preparation andprovision of the early warning message willlead to improved reaction capabilities of theaffected people.

ReferencesPetley, D.N., Higuchi, T., Petley, D.J., Bulmer,M.H. And Carey, J., 2005. Development ofprogressive landslide failure in cohesive mate-rials. Geology 33(3): 201–204.

Subproject: Coordination, Integrationand Optimisation of a Multi-SensorSystem for Monitoring of Landslides

Rolf Becker, IMKO Micromodultechnik,Ettlingen

The most important objective of the »SetupMonitoring« project within the collaborativeresearch project is the coordination and hard-ware-related integration of heterogeneousfield sensors into a unified, robust and simpleto use measurement system, as a standalonecomponent of an extensive landslides earlywarning system.The innovation of the monitoring system areits particular measuring procedures and inte-grated data recording mechanisms, which willbe specifically adapted to the early warning ofrotational slides. The system includes sensors

for determining the load (meteorologicaldata), the inner status of the vadose zone(water content, soil suction head, and porewater pressure) and the system response(assessed by monitoring kinematics usingGPS, tilt meters, inclinometers, and exten-someters) for a slipping body. Conventionalsensors as well as novel measurement proce-dures will be used.The time variant soil water dynamics is the keyfactor ruling the current geo-mechanical sta-bility of a slipping body. Determining the soilmoisture in clay or highly electrically conduc-tive soils is a technological challenge due toenergy dissipation during the measuring pro-cedure. The measuring principle Time-Domain-Reflectometry (TDR) is less prone to theseeffects and thus especially suited for the par-ticular application. TRIME-TDR sensors byIMKO supply reliable measurements even withdifficult soils, and are therefore used as thesolid backbone of ground water measure-ments related to landslides.The novel stationary geoelectric system fromthe project partner geoFact is one of the firstbeing capable of continuous monitoring andwill be a key component of the integratedmonitoring system. Scale transition from pointto slipping body is achieved by the extendedtwo- or three-dimensional conductivity fieldsresulting from geoelectricity in combinationwith the pointwise soil moisture measure-ments used for calibration. The subproject willsupport geoFact in developing a procedure forupscaling soil moisture.The »Spatial TDR« method, currently beingdeveloped at several german universities andresearch institutes, allows the determininationof water content profiles along elongatedsensor cables of several meters length. How-ever, this procedure requires a large mathe-matical effort to analyse signals and locallydoes not achieve the same accuracy as con-ventional TDR sensors. As part of the projectit will be tested whether a combination ofSpatial TDR and standard TDR sensors pro-vides a significant information gain concern-ing infiltration, purched ground water tables,and hanging slippage.

93

Another forward-looking aspect of the systemintegration is the fusion of IMKO’s well estab-lished sensor technology for recording envi-ronmental variables with novel self-organisingwireless networks, which will be installed andoperated by the subproject »SensorGIS«. Asmall number of the sensors for vadose zonemonitoring will be taken out of the previouslybuilt cabled field bus system and will be inte-grated into the wireless network. The wirelessnetwork nodes from ScatterWeb have a vari-ety of interfaces to connect the sensors. Heretoo, hard- and software adjustments will prob-ably be required to match the different inter-faces. Robustness, prevention of downtime,and energy supply for the planned sensor net-work are important aspects of the joint inves-tigations. These issues are decisive criteria forfuture applications of wireless sensor networksfor environmental monitoring.

Subproject: Cooperative RiskCommunication

Marco Danscheid & Jürgen Pohl,University of Bonn

The subproject »Communication« has twosuperordinate aims: One is the clarification ofgeneral local and regional needs. The otherregards the cooperative implementation ofearly warning systems with both the affectedplayers and the other subprojects.As a logicalstarting point the work is oriented on the endusers needs. These shall be assessed in detailright from the beginning of the project andthey shall influence the design of the EarlyWarning System that is to be developed. It isof decisive importance to identify the infor-mation required and how it should be pre-sented. More pointedly, do the end users wanta »flashing red light« or do they prefer moredetailed information, which require them tocome to a decision?Questions are: Which protection-worthygoods exist on the localities? Which con-crete need for action does that imply?Which kind of early warning information dothe players need?

Where men are involved, one can speak ofsocial systems. Every social system (e.g. com-panies or public authorities) has specific logicsand connected languages. To investigate theseis an important aim of this project.Since there are different logics and lan-guages in different social systems, it isimportant to take this into account for riskcommunication. It is essential to developshowcase translations to enable differentsocial systems to communicate.Questions are: How can be guaranteed thattimely communicated early warnings will betransformed to adequate reactions? How toconceive an early warning system that isaccepted by the end users?Examining other research projects on »earlywarning of landslides« it becomes obviousthat social scientific components are eithercompletely neglected or only implemented asa simple communication module. These mod-ules are often only used to communicate nat-ural scientific results to the players. A cooper-ative implementation of the Early Warning Sys-tem, where players – both end user and con-sultant – are taken seriously right from thebeginning does not exist, at least not as it isspecified in the plans for the subproject»Communication«.The subproject »Communication« would liketo break new ground by penetrating the com-plexity of social actor systems with the help ofcooperative interviews and by developing sen-sible solutions in collaboration with the play-ers. The aim of this approach is to make a con-tribution to an early warning system whicheven works in the ultimate consequence – andin doing so saves lifes.The cooperative implementation of the EarlyWarning System stretches across the wholeproject time. A main focus of the subproject»Communication« is to do qualitative inter-views with the involved players and to analysethe findings.Based on the analyses showcase translationschemas are developed. They shall help toimprove the communication between differentactor systems or even to enable communica-tion for the first time. The question is, how

94

does the communication of a player have tobe formed for other players to perceive it andacknowledge it as relevant.

Subproject: Historical comparativeregional analysis of frequencyand magnitude of landslides

Andreas Dix & Matthias Röhrs,University of Bamberg

The subproject’s objective is to develop meth-ods for monitoring frequency and magnitudeof landslides through history. This is to be con-ducted based on research of the Swabian Albregion and compared with South Tyrol in twolandslide-prone regions. Based on the resultsand experiences gained, conclusions shall bemade as to how historical analysis can in thefuture be integrated into an early warning sys-tem which is as effective as possible.Historical data plays a decisive role in the com-plex chain of early warning and risk commu-nication in at least two system areas:1. Current knowledge of the spatiotemporal

distribution of past landslide events is veryincomplete. The effectiveness of an earlywarning system, however, depends uponthe quality of data especially of the fre-quency and magnitude of events. It is the-refore necessary to use all existing datapools to implement an early warningsystem, which includes the historical infor-mation saved in the archives. Systematicresearch and indexing of the historicalmaterial thus helps to improve the knowl-edge of the total distribution of landslides.Only by establishing event series whose spa-tial and temporal resolution is as high aspossible, can conclusions regarding riskzones and expected distributions of futureevents be made. Along with the measure-ments of current conditions, these form thebasis of a significant early warning system.This data can also be used to check existingdata for its representativeness and validity.

2. Apart from scientific risk analysis, an earlywarning system is only effective if it is basedon a high public risk awareness. This knowl-

edge, on the other hand, can essentiallyonly be based on past events. Thereforehistorical knowledge seems fundamentalfor the successful implementation of anearly warning system. This knowledge alsoincludes handing down perception andhandling of these specific natural hazardsthrough history. The question of what andhow much of this knowledge was actuallyhanded down, can be adequately recon-structed by analysing the archives.

Data acquisition is to be carried out compara-tively in cooperation with the subprojects fromthe monitoring, modelling and implementingclusters in two sufficiently different testregions, the Swabian Alb and South Tyrol.The first work step, the index books of cen-tral archives are to be analysed so that thefollowing step the relevant files can besearched through thoroughly and other localarchives included.This should allow long dataseries regarding frequency, magnitude, trig-gering factors and perception of landslides tobe created.To achieve results in a short period of time thesnowball effect has proven successful for pre-vious investigations. This does not only takeinto account all available types of source butalso keeps track of any considerations asregards other subprojects, experts or the localpopulation. Here an exchange in both direc-tions has proven to be fruitful.It can be assumed that based on the experi-ences in the Swabian Alb region systematicresearch not previously carried out in SouthTyrol will produce similar results. The sourcedensity in South Tyrol is supported due to thefact that from the 19th century the HabsburgMonarchy conducted high-resolution cadas-tral mappings precise by lot which can berated as very good in terms of quality anddensity. We also assume that during theAlpine war from 1915 not only measurementsbut also the first aerial view series and pho-togrammetric measurements were conductedon a large scale which had never been usedfor such purposes before.

95

Subproject: Integration of early warninginto an integrated risk management

Stefan Greiving, plan + risk consult,Dortmund

An early warning system has to be developedaccording to the requirements of its users whohave been identified in the Cluster »Imple-mentation«. Therefore, the management ofinformation as well as the dissemination ofinformation (risk communication) are of par-ticular importance. Thus, there is a strong con-nection to both of the other sub-projects thatbelong to the cluster »Implementation«.Research objective of the part »Manage-ment«, to be awarded by contract, is tobroaden the perspective and to provide thestakeholders in the case study areas with anappropriate consideration of action alterna-tives. An early warning system however is onlyone of many alternatives because the whole»disaster management cycle« of prevention,preparedness, reaction and reconstruction hasto be considered. An early warning systemthus may not have to be understood as a sin-gle isolated measure of which the implemen-tation is based only on the identified hazard.Moreover, existing vulnerabilities and actionalternatives have to be considered, too.As a part of the envisaged integratedapproach, different and sometimes alternativemeasures may compete with each other. Cri-teria for the assessment of measures – thathave been defined in co-operation with localstakeholders – are especially the protectiongoals for hazard-prone areas but also aspectsof efficiency and effectiveness.Consequently, this part contributes to improvethe implementation of the early warning sys-tem to be developed and tested as well as toimprove the economic exploitation by theinvolved companies. Further, it will contributeto a secure and efficient use of public fundsthat are needed in order to establish earlywarning systems in practice.The existing fragmentation of isolatedapproaches often derives from the broad dis-tribution of responsibilities for action within

the disaster management cycle among numer-ous persons and/or institutions in charge.According to the fact that planning and imple-mentation competences are only bundled atthe local level and taking into account that theproximity to potentially affected peoplepromises a high level of involvement of con-cerned stakeholders, the focus of the workwithin this part will be on the local authoritiesof Lichtenstein-Unterhausen (Swabian Alb,Germany) and Nals/Nalles (South Tyrol, Italy).The relevant stakeholders in the Swabian Alb(association of administrations, county, Re-gional Association Neckar-Alb) and in SouthTyrol (Autonomous Province of Bozen/Bolzanowith its responsible departments) will be iden-tified and involved in the project.

Subproject: Central Spatial DataInfrastructure, Open Web Servicesand Web Processing Servicesfor the Development of an Informationand Decision-Support System for RiskManagement in Early Warning Systemsfor Landslides

Klaus Greve, University of Bonn

This subproject aims to develop the basic tech-nological principles for an early-warning spe-cific spatial data infrastructure. Early warningsystems require very complex information andcommunication structures. They need to inter-face sensor and data management technologywith database management systems, datapreparation and analysis, prognosis and modelcalculations, computer networks, communica-tion terminal units, result processing systems,and communication tools.To date, major problems on syntactic andsemantic levels have resulted from the con-nection of heterogeneous components incomplex Systems. Basically, such interconnec-tions are easily controllable by the use ofclassical technology. However, due to thespecial complexity of early warning systems,the coupling of numerous components andinterfaces (syntactic problem), differentrequirements on the content (semantic prob-

96

lem), and a highly dynamic technology devel-opment, the interconnection is more difficultand potentially lossy. In these cases, infra-structure oriented technologies described bythe keywords Sensor Web, Open Web Ser-vices and WPS will be used. They are basedon the Web Service Paradigm and originatedin context of Spatial Web approach of theOGC. This technology offers much higherinterconnection and networking abilities thanclassical technology.Web Services isolate different ›jobs‹ of spa-tial information processing and implementthem in the architecture of different mod-ules. Theses modules can be implemented asstand alone services and communicated viastandardized interfaces. Open Web Services(OWS) are the basis of spatial data infra-structure projects on international (GSDI)and national levels (GDI.de), and in nearly allof the German federal states (GDI NRW, GDIBB, GDI NI etc.).Initial pilot tests have indicated that the con-cept has reached a level of possible ›realworld‹ implementation. Part of the informa-tion provided by these services includes assess-ments about the service’s own ability to pro-vide information. Interlinking is not only possi-ble along linear structures, but also withincomplex and partly self-organizing networks.These services thus already anticipate structur-al elements of the Semantic Web, the succes-sor to the WWW.In this subproject, the spatial information tech-nology basis of early warning systems will beexplored by first implementing current con-cepts of geoinformatics into the service archi-tecture and processing modules. The strengthsand weaknesses of the solutions will then beevaluated within the application. In order todo this, the central components of an earlywarning specific spatial data infrastructure willbe implemented.The basic concept and all essential modules ofthis subproject will be developed and imple-mented as WebServices, and therefore usefulin other Spatial Data Infrastructures.The concept is considered to be innovative,and real-world efficiency can be supposed.

In this project, a high complex Spatial DataInfrastructure with very heterogeneous infor-mation sources and information will be creat-ed, as defined by the involved organisations.

Subproject: Development of an adequatedata model schema for an informationand decision support system for riskmanagement in landslides early warningsystems

Stefan Jäger, geomer, Heidelberg

A standardized early warning process for massmovements requires a well structured infor-mation management for all relevant informa-tion and processes. This sub-project is a crosscutting issue for the project’s framework out-line. The temporal and spatial uncertainties onone hand as well as the great variety in qual-ity, amount and availability of relevant data tobe expected on the other hand, require com-plex database schemas. These must have theability to completely represent the early warn-ing chain and thus the activities of the moni-toring and the modeling clusters of ILEWS. Theinformation and decision support system musthave the ability to help decision makers as wellas disaster management with qualified deci-sion support. This includes also informationconcerning the triggering mechanisms of land-slide processes and their temporal and spatialquality, which is dealt with in the modelingcluster. That means, information must be cov-ered and conveyed ranging from dense, site-specific monitoring systems on single landslidebodies, to simple landslide susceptibility mapsand statistically poorly sustained rainfall inten-sity/duration triggering indices. The oftenuncertain information situation also requiresthe possibility to represent conclusions basedon probabilities. In addition, a landslide earlywarning system must be able to provideestimates of the potentially associated risksfor the affected population and the econom-ic activities. Information about critical infra-structure locations and their importanceare therefore an essential part of the infor-mation process.

97

– Definition of a geodata model for the rele-vant information components. The geodatamodel consists of basic geodata as wellas of monitoring data from sensor webs,socio demographic data and critical infra-structure information, with respect to OGC-conformal data formats and description(metadata standards).

– Definition of interfaces to already existingdata infrastructures of the national andmunicipal civil defense authorities, underconsideration of interoperability standards ofthe OGC. Here, the task is to develop web-based services for the dissemination of thedata the data and information sampled andcompiled by the sub-projects to the users(civil protection, decision makers etc.).

– Development of an AJAX based visualizingcomponent for various user groups and devi-ces on the basis of the before mentionedweb services. This is of concern primarily forthe monitoring and measuring of data,which serve as input for models and for themodel results themselves. These data shouldbe visualized preferably in real time and forvarious time steps for visual control. Thevisualization component mainly will bedeveloped based on standardized OGC com-patible mapping services as far as geodataare concerned. Additional developmentdemand is given in the field of visualizationof the historical data as well as for the socio-demographic and economic data in the con-text of early warning.

– Implemantion and tests of the systemscomponents will finalize the develop-ment process.

Subproject: Geodetic Monitoringand Modelling

Heiner Kuhlmann & Stephan Burghaus,University of Bonn

The task of geodetic monitoring measure-ments is to get a confirmation of predictablechanges (e.g. subsidence behaviour of build-ings) or the proof for a non-expected or non-predictable change of an object (e.g. land-

slide). Information is generally being suppliedthrough selected measuring points. Thebehaviour of the object can be quantified byanalysing the movements over time. Closelyrelated with the determination of movementsis also the question of reasons in order toderive a causal connection.By means of a specially created geodetic pointnetwork that spreads over stable as well ascritical slope areas those areas should be iden-tified whose movement intervals differ signifi-cantly from other areas due to certain othereffects (e.g. increase of humidity, change ofpore water pressure, etc.). Absolute move-ments of ground points in slope areas arebeing recorded and compared to referencepoints via measuring methods such as GPSand electronic precision tacheometry. Apartfrom those (geodetic) network points furthermeasuring stations are being created whichare equipped with sensors for relative meas-urements (e.g. chain inclination measuringsystems in the subproject »Geomorphic Mod-elling«). They must be linked with the geo-detic measuring points in order to get bestredundant but assignable measuring informa-tion on movements.It is the intention to use both geodetic meas-uring methods one after another. If weassume that the movement intervals are about0.3 mm/month, the measuring resolutionof the precision tacheometer of about0,2–0,3 mm will be sufficient if you carry outepisodic measurements and repeated meas-urements about every 2–3 months. They donot only cover the monitoring of the geodeticpoint network but also the respective integra-tion of all measuring stations for relative meas-urements. This does not only deliver redun-dant information on areas close to eachother but the automatic relative measure-ments can serve as indicators for beginningmovements in order to possibly initiate mon-itoring measurements outside the scheduledmeasuring epochs.As already indicated GPS measurements shallbe carried out with the same measuringepoch. They do advantages with continuousmonitoring methods over several weeks

98

provided they are being carried out as staticmeasurements. The data achieved in thelocal network come together in a centralcontrol and evaluation unit, by that it ispossible to calculate the basic lines be-tween the network points automatically andcontinuously in order to derive station move-ments from that.The observation deviations with the GPSmethod show a certain auto-correlation inthe range of a few minutes up to a fewhours. Reasons for this are e.g. multipath andextension effects of the electromagneticwaves. The dimension of the deviations liesabove the point movements to be expected.A reduction of those observation deviationscan be achieved during long observationperiods and an analysis using the post-pro-cessing method. Due to the planned earlywarning system a real-time process is wantedhere which arranges the analysis in such away that a separation of measurement devi-ations and point movements will be done ina filter approach.To get a relative precision of tacheometricmeasurement clearly less than 1 ppm a regu-lar examination of the measuring instrumentwill be necessary. It is also not sufficient tointroduce the gained meteorological param-eters of the endpoint as representative factorsfor the entire measuring distance. Hence thedetermination of the refractive index plays animportant role and finally is the precision lim-iting factor for the distance measuring espe-cially in mountained areas.A successful way for high-precision distancemeasuring was taken in the 1990s. Basedupon the light dispersion in a turbulent medi-um, the fluctuations due to atmosphericexchange processes are described in a modelby means of statistic factors. Suitable com-mercial systems to measure these atmosphericfluctuations have been developed by Scin-tec/Tübingen in form of the scintillometermeasuring systems. Regarding the chosenstudy areas a scintillometer will be used, whichcan do measurements up to 4–5 km. Its usagewill lead to a significant improvement with themodelling of the refractive index and can

therefore cover huge parts of the refractivecomponents which have previously been diffi-cult to determine.In the chosen study areas changes are beingexpected that have an explicit time connec-tion. The movements to be achieved can bemodelled together as a function of rainfall,pore water pressure, slope inclination andground parameters, etc. This will then be thebasis for an Early Warning System via which anaccording measure and emergency conceptshould be initiated in order to best handle thecurrent situation.

Subproject: Spatial Monitoring of soilparameters with geophysical surveymethods

Heinrich Krummel, geoFact, Bonn

Primary objective– The aim of the project is the development of

an innovative monitoring system for soilparameters and flow potential based ongeophysical survey arrays.

The intention is to permanently install sturdy2D/3D geoelectrical survey systems in poten-tially endangered landslide areas. An automat-ic procedure will be developed for data collec-tion and for transferring the data via modemto a central processing unit. The calibration ofthe geoelectrical data is done by singular insitu soil moisture measurements (co-operator:IMKO, Karlsruhe). The processed results of thesurvey will be implemented into a centraldatabase for the mutual use of all cooperatorsinvolved in the main project and can be eval-uated and interpreted with respect to thecommon goal of the development of an earlywarning system for landslides.Within this sub project it is intended to per-manently install a geoelectrical survey array ata known landslide endangered location inLichtenstein-Unterhausen (Schwaebische Alb,Germany). The landslide body will be exam-ined with geophysical methods (e.g. seismic,geoelectric) prior to the installation to deter-mine the geometry of the body and to deter-mine critical »hot spots« for the setup of the

99

permanent array. The survey system will auto-matically perform several surveys each day.Special (TDR-)probes of partner IMKO GmbHwill be installed within the array at differentpositions in different depths to measure soilmoisture. The geoelectrical results will be cali-brated using the probes to gain areas of dif-ferent soil moisture out of the geoelectricalresistivity data.Automatically procedures for data collection,data transfer and analysis will be developedwithin the sub project.

Scientific and technical objectivesOptimisation of 2D/3D-geoelectrical surveyarrays (e.g. linear, star shaped or quadraticarrays, electrode separation) for continuoussurveying of soil moisture/flow potential at»hot spots«, e.g. sliding surface of a potentiallandslide.– Development of an automatic procedure for

periodic data collection and data transfer– Development of an automatic procedure for

analysis of geoelctrical data for interpreta-tion of soil moisture conditions, therefore:

– Development of an automatic inversion pro-cedure for creating to create geoelectricalresistivity models from apparent resistivitydata.

– Scale transformation for soil moisture fromsingular in situ measurement to spatial infor-mation by using 2D/3D-geoelectrical surveysystems.

– Development of a prototype of an automa-ted 2D/3D-geoelectrical survey system formonitoring the soil moisture/flow potentialat potential landslide locations

Subproject: Standardised, wireless sensornetworks for the efficient acquisition,transmission, storage and visualisationof geodata

Hinrich Paulsen, terrestris, Bonn

Warnings about imminent events in the realmof mass movements require timely informationabout certain parameters. Amongst these arechanges in slope angles, pore water pressure,

precipitation intensities, soil moisture contentsto name but a few. The aim of terrestris is toemploy wireless sensor networks (WSN) toefficiently acquire, transmit, store and visualiserelevant geo-data.Data will be visualised in a web based geo-graphical information system (WebGIS). Inevery phase of the project the use of aWebGIS has the added advantage of beingable to immediately evaluate and quality checkincoming data with ensuing modifications tothe system gradually optimising the sensors.Following this procedure it is possible to eval-uate the robustness and functionality of thesystem – on the one hand with regard to thedata and on the other hand with regard to thehardware. Failed sensors due to lacking ener-gy, frost or mechanical damage, etc. willimmediately show up in the WebGIS.Apart from the above mentioned advantagesthe WebGIS can also function as a communi-cation platform which is particularly usefulwhen dealing with spatial data since it alsofacilitates the coordination of and discussionamongst spatially distributed project partici-pants. Other players like communities, scien-tists working on the same test site, or even thegeneral public can easily be granted access tothe system.Wireless sensor networks are a new type ofgeographical information system for in-depthand continuous monitoring of the environ-ment. Technically termed »embedded sys-tems« these mini computers were only possi-ble through the progress in semi-conductortechnology of recent years. They consist of acentral processing unit (CPU = microcon-troller), memory and radio technology. Thisbasic hardware can be equipped with sensorsof any kind like those measuring temperature,vibration, movement, humidity, etc. On princi-ple all sensors can be attached that provide anelectronic signal.Due to the fact that sensors are distributed inspace they can be assigned a coordinate whichturns the measured data into geodata. Thisgeodata is then routed to a special node withaccess to the internet by means of the GlobalSystem for Mobile Communications (GSM),

100

General Paket Radio Service (GPRS) or radioand is directly written to a database. Thesedatabases are deemed to be object-relationalwith some of them (PostgreSQL, Oracle) beingable to directly store geographical featuresthrough the use of a spatial extension.Collaborative geographical information sys-tems are dependent on standards to functionefficiently. The Open Geospatial Consortium,Inc. (OGC) works with government, privateindustry, and academia to create open andextensible software application programminginterfaces for geographic information systems(GIS) and other mainstream technologies. Oneof its initiatives is the Sensor Web Enablement(SWE) activity which is establishing interfacesand protocols that will enable sensors of alltypes to be accessed over the Web.

101

Development and testing of an integrative 3Dearly warning system for alpine instable slopes(alpEWAS)

1. IntroductionAlthough great advances in the recognition,prediction and mitigation of landslides havebeen made in the last few years, major eventsespecially in alpine regions still claim a highsocial and economical tribute. Especiallythrough extreme weather conditions, as e.g.the intense rainfall in August 2005, instableslopes can be activated and endanger people,settlements and goods in its surrounding. Cur-rently an increase of this problem caused bythe global climate change can be observed.Recent landslides, which occurred in the alpineregion, demonstrate the need for a deeperunderstanding of the geological and physicalprocesses, which can lead to a spontaneousfailure of a natural slope. Major rockslides asVajont (1963, Italy) or Randa (1991, Switzer-land) and recent minor events as Sibratsgfäll(1999, Austria) prove the destructive potentialof these mass movements and the need toinvestigate the mechanics of such processesmore deeply. Progress in the assessment of theland slide risk will only be achieved if the trig-gering processes and the kinematics of themovements are better understood.To accomplish this task an assumedly instableslope has to be examined for its engineeringgeological properties and then has to beobserved continuously with a suitable moni-toring system. Exclusive instruments and

methods to achieve this are available, but foreconomical reasons they are rarely used. Atthe same time the number of localities withneed for monitoring is rising noticeably.Therefore the goal of the alpEWAS joint pro-ject, which is being carried out by the Tech-nische Universität München and the Univer-sität der Bundeswehr München, is to developand test a relatively economic and widelyapplicable monitoring and early warning sys-tem at a designated location.The monitoring system is based on the inte-gration of innovative and economical measur-ing technologies to a Geo-Sensor Network.The surface movements will be detected punc-tiform and highly precise with the Global Nav-igation Satellite System (GNSS) as well asextensively in a large part of the landslide areathrough reflectorless tacheometry; the meas-urement of the movements in the depthalongside boreholes will be done by using anewly adapted Time-Domain-Reflectometry(TDR) System. Parallel to theses measurementsthe registration of the hydrostatic pore pres-sure, as well as the climatic conditions at thesite will be carried out. In this way the 3Dmovements of the slope, which are deter-mined nearly in real time, can be comparedwith the surrounding conditions (precipitation,hydrostatic pore pressure etc.) and can beanalysed for trigger mechanisms.

Thuro K. (1), Wunderlich T. (2) & Heunecke O. (3)

(1) Prof. Dr. Kurosch Thuro, Lehrstuhl für Ingenieurgeologie, Technische Universität München, Arcisstraße 21,

80333 München, [email protected]; http://www.geo.tum.de.

(2) Prof. Dr. Thomas Wunderlich, Lehrstuhl für Geodäsie, Technische Universität München, Arcisstraße 21,

80333 München, [email protected]; http://www.geo.bv.tum.de.

(3) Prof. Dr. Otto Heunecke, Institut für Geodäsie, Universität der Bundeswehr München, 85577 Neubiberg,

[email protected]; http://www.unibw.de/ifg.

102

Through the large amount of data collectedby the system (in time and space) it shouldbe possible to determine causal and tempo-ral coherences between the most importantinfluencing factors and the movement of theinstable slope within a relatively short periodof time (6 to 9 months), allowing the defini-tion of critical threshold values. By using anautomated alarm function, which informs theresponsible authorities when the threshold isexceeded, an early warning system can beimplemented. A major task of the alpEWASproject is to develop the necessary hard-and software and to test the system at adesignated site.A real cost advantage of the readily devel-oped monitoring system is achieved by mak-ing remote maintenance and inquiry possible.All important functions can be queried andcontrolled from the project office via Modem,reducing the costs for personnel in compari-son to conventional measurements by elimi-nating the need for repeated manual meas-urements. Furthermore the installation costs

are – when compared to the amount of datawhich is gathered in space and time – rela-tively low due to the used measuring tech-nologies (reflectorless tacheometry, low-costGPS and TDR). The linking of the differentmeasuring units with the local control centerwill as far as possible be accomplished byusing WLAN, which makes the usage of anexpensive mobile telephone system or beamradio unnecessary.

2. Project SiteThe landslide »Aggenalm«, situated in theSudelfeld area near Bayrischzell (Bavaria), hasbeen selected as the location for the installa-tion of the early warning system (Figure 1).After a first engineering geological investiga-tion, the location has been proven suitabledue to its movement rates of about 2 cm peryear and the anticipated maximum depth ofthe landslide of about 25 m in most areas.Additionally the Aggenalm, a grass slope inter-spersed by single rock blocks, is ideal for thedevelopment and training of the new moni-

Figure 1: Orthophoto (scale ca. 1 : 5000) of the landslide Aggenalm with displacement vectors in the scale ca. 1 : 1(observation period: 2003–2004). The shown results emanate from periodic geodetic measurements of the slope whichwere carried out on behalf of the Bavarian state office for the environment (Bayerisches Landesamt für Umwelt). In theupper half of the picture a secondary landslide which occurred in 1997 can be seen. The Aggenalm and the accessroad to the ski resort Sudelfeld are affected by the movement.

103

toring strategies needed when using reflector-less robot.

3. Deformation measurement techniques

3.1. Time Domain Reflectometry (TDR)Time Domain Reflectometry (TDR) is widelyknown as a system for the measurement ofsoil moisture (e.g. TOPP, DAVIS & ANNAN1980). With few modifications TDR can alsobe used for the monitoring of localized defor-mation in rock and soil. To date this applica-tion has only found wider acceptance in NorthAmerica, while it is still largely unknown inEurope. This is surely based on the fact that sofar most of the research has been carried outat the Northwestern University (Evanston/Chi-cago, Illinois) under the leadership of Dowdingand O’Connor, who have without doubtproved the usability of TDR in landslide moni-toring (O’CONNOR & DOWDING, 1999).Especially the comparably low installationcosts and the possibility to perform continuousmeasurements make TDR an interesting alter-native to inclinometers. Presently, TDR land-slide monitoring systems are capable of deter-mining the exact depth of the observed defor-mation zone, while only a semi quantitativestatement can be made of the amount ofmovement. The orientation of the movementcan not be determined at all. Furthermore inmost instances the application is limited tothe measurement of localized deformationas it is typically observed in rock (e.g. localiz-ed shearing alongside joints) (KANE, BECK &HUGHES, 2001).In the opinion of the authors some of thesedisadvantages can be overcome by definingstandardized installation procedures (e.g.grout type, coaxial cable type) adjusted fordifferent geologic settings. Furthermore newmethods for the analysis of the received TDRdata are under development, especiallywhen multiple TDR measuring points areconnected to produce a 3D model of thedeformation zone.The monitoring of deformation in rock/soilusing TDR is based on an indirect measuringmethod: the deformation itself is not mea-

sured (as with an inclinometer) but a directlydependant value, the change in impedance ofa coaxial cable as a result of deformation, ismeasured.TDR can be described as »cable-based radar«and consists of two basic components: a com-bined transmitter/receiver (TDR cable tester)and a coaxial cable (Figure 2). The TDR cabletester produces electric impulses, which aresent down the coaxial cable. When these puls-es approach a deformed portion of the coax-ial cable an electric pulse is reflected and sentback to the TDR Cable Tester. The reflectedsignals are collected and analysed. As withradar, by measuring the time span betweenemission and reception of the signal the dis-tance between the TDR Cable Tester and thedeformation can be determined. Furthermoreby analysing the reflected pulse (amplitude,width, form etc.) information about the typeand amount of deformation can be obtained.For landslide monitoring a semi rigid coaxialcable is installed into a borehole and connect-ed to the rock mass with grout. There are basi-cally three different installation methods(Figure 3): 1. The TDR cable is installed paral-lel to an inclinometer within the same bore-hole; 2. the coaxial cable is installed into asheared inclinometer casing, therefore extend-ing the lifespan of an inclinometer borehole;3. the coaxial cable is installed into a boreholeof its own.The installation parallel to an inclinometer inthe same borehole is primarily for researchpurposes, since a direct comparison of incli-nometer measurements (direct measuringmethod) with the TDR readings (indirectmeasuring method) is made possible. For thisreason this method will be an important partof the ongoing research. Neverthelessoptions 2 and 3 are probably the most appro-priate installation methods for commercialuse: option 2 because it extends the lifespanof an existing inclinometer borehole, andoption 3 because a sole TDR installation canbe established at comparably low costs,because of the relative small boring diame-ters required and the low cost of the coaxialcable. Furthermore with option 3 the grout

104

Figure 2: Basic setup of a TDR measuring site. The coaxial cable is installed into an instable slope and connected to theTDR cable tester. As soon as the coaxial cable is deformed by the mass movement a peak can be seen in the reflectedsignal. Its amplitude is dependant to the amount of deformation taking place

Figure 3: Possible installation setups for a TDR coaxial cable into a borehole with and without an inclinometer

105

and other installation parameters can beoptimized to greatly enhance the perfor-mance of the TDR measurements.Many different factors influence the measure-ment of deformation with TDR (O’CONNOR &DOWDING, 1999). Generally two groups ofeffects can be distinguished: the influence ofthe deformation itself and the installationparameters of the TDR system.TDR measurements are influenced greatly bychanges in the geometry of the measurementcable. The type (shearing, extension and com-pression) and the amount of deformation, aswell as the width of the deformation zone,affect the received signals. Prliminary simpletests have shown that discerning the differenttypes of deformation is not always possible.However, especially with progressive deforma-tion, a determination of the deformation typeis sometimes possible.It is evident that the components used in aTDR system, and the way it is installed into alandslide, influence the measurement. In par-ticular, two parameters stand out in theirimportance: the type of coaxial cable usedand the physical properties of the grout. Thematerial from which the cable is made, itsdiameter and the length all influence themeasurements. Generally, when using thickercables, larger total deformations can beobserved before the cable is severed. At thesame time, the sensitivity of the system tosmall movements is reduced. Accordingly,thinner cables should be used for landslideswith low deformation rates, and thickercables for »faster« landslides.The grout is the interface between the cableand the rock mass. It is, therefore, importantto match the physical properties of the groutto the particular geology, especially whenworking in soil. If, for example, the grout istoo strong, a pillar-effect might occur where-by soil moves around a pillar over-stiff grout,resulting in too little or no deformation of theTDR cable.Varying the type of cable and the grout mix tosuit the surrounding geology and the antici-pated deformation rates will enhance thequality of the received TDR measurements.

This is one main task of the ongoing research:defining certain installation procedures for dif-ferent geological settings and deformationrates. Furthermore a careful calibration ofthese setups is a pre-requisite for ensuring themaximum quantity and quality of information(especially the deformation type and amount)through the indirect measurement of defor-mation with a TDR system.Within the alpEWAS project these newlydefined installation setups, which were deter-mined in laboratory shear tests, will be testedfor the first time in the field. Signal analysissoftware will be developed, which will allowan automated processing of the TDR signals.The TDR measurements are then comparedand if necessary further calibrated with paral-lel inclinometer measurements.

3.2. Reflectorless Robot-TacheometryThe component RL-TPS (Prismless TerrestrialPositioning System) fulfills the task of interpo-lation within the joint project. Its aim is to den-sify the displacement pattern determined byTDR and GNSS at only few selected loci andfocusses on discovering local maxima or sin-gularities. To attain high economic efficiency itshould be avoided to fit targets with specialprisms. An exception is only allowed for thosepoints which are needed for continuous test-ing the stability of the observation stationitself and to confirm the motions of the GNSSand TDR spots. Most objects to observe will benatural targets, e.g. rocks standing out of theslope and tree stumps.Spatial polar methods based on automatic tar-geting robot-tacheometres with prismlessEDM are well suited for periodical surveying ofthese objects. By corresponding programmingthe geodetic precision instruments are able toscan suitable objects with a dense grid. The3d-model derived from this grid can be inves-tigated for displacement and deformation bycomparison with a 3d-template from the ini-tial epoch. The 3d-templates themselves shallbe determined in close-range manner by a ter-restrial laser scanner with high resolution. Todetermine the spatial components of themovement, corresponding software has to be

106

developed based on known geodetic algo-rithms. In addition corresponding statisticaltests of significance have to be provided.The solution could be substantially improved,if it were possible to make use of the internalimage information collected by the automatictarget recognition unit. In this way contourscould be extracted and possibly grey valuescould be compared. The Chair of Geodesy dis-poses of an instrument with such an interfaceand pioneered successfully automatic pointingat natural targets. In spite of this the fix focusfeature of the internal camera prevents theapplication for the present problem, becauseless than 500 m distance images will be poor-ly defined. A single manufacturer has devel-oped a prototype of a focusable internal cam-era; although it is not yet on the market, anoperative instrument based on the actualtacheometry technology was ordered by theChair of Geodesy. First tests and calibrationprocedures already are planned and must bescheduled before the system can get operativeat the site.These additional tasks must be performed toreach the highly innovative final aim: to real-ize the vision of a system, able to auto-nomously find and select suitable natural tar-gets. Only with the image information it willbecome possible to train a survey robot tosearch systematically suitable natural targetson a slope, to scan them and to subsequentlyperform a periodical motion check. The nec-essary strategies have to be developed andprogrammed in a way that enables the RL-TPSto adapt itself for changing atmospheric andtopographical conditions. The accuracy, thepromptness and the reliability of the RL-TPSwill be slightly lower than TDR and GNSS asthe sophisticated processes of observation andevaluation will lead to a certain delay withrespect to the numerous natural targets.Another reason comes from a necessary gen-eralization which cannot be prevented duringthe modelling routines. Nevertheless, in totalthe evidence concerning the entire behaviourof the landslide will be of substantiallyincreased value in order to gain insight intothe complex kinematics.

The high economic advantage of the strategyto use natural objects as targets is opposed byan unavoidable disadvantage: especially in thealps those targets will be covered with snowduring winter season and therefore will beinvisible. Moreover, it has to be expected thatdue to the cold weather the supply of solarenergy will be weak. Therefore we suggest tolet the TPS scan the current snow cover oncea week. This information can be used to com-pute the water volume from melting snow inspring to incorporate the most efficient land-slide trigger into the general model.Low-cost GNSS and TDR are very suitablesolutions to continuously gather highly pre-cise information about the movements onthe surface and underground, respectively. Bydoing so, the tendencies at a few selectedpoints within an instable slope can be detect-ed economically. Despite using great careful-ness to select representative sites for themeasurements, they still are only spot checks,which can miss areas with high rates ofmovement and therefore make a deeperunderstanding of the underlying landslidemechanism difficult.Thus it is necessary to economically gain infor-mation covering the complete landslide area,in order to incorporate all problematic regionsinto the monitoring system by either movingexisting sensors or installing new ones. Whenhigh rates of movement are encountered, themethods of aerophotogrammetry and satelliteremote sensing are used. At lower rates ofmovement and the need for continuouslyupdated information, TPS is to prefer for eco-nomical an accuracy reasons. However, adense placement of reflectors within an insta-ble slope is not reasonable. The snow depthswhich have to be expected in the mountainsforce the installation on long metal posts.These would not only deface the slope andproduce an unwanted psychological sensationof threat, but with time would also becomeaskew and attract lightning during thunder-storms. Additionally the installation of theposts would lead to extra costs, especiallywhen using active reflectors (dependent onmanufacturer), which need an extra energy

107

supply. Thus an economic solution must aim atnatural targets. This has already been triedusing reflectorless tacheometres, which sweepthe terrain in a manner comparable to a scan-ner, but did not lead to a verifiable successwhen used for the monitoring of instableslopes – other as when used for the observa-tion of snow accumulations for the mitigationof avalanches (SCHEIKL et al. 2001).The new approach proposed in this project isnow trying to combine the advantages of twoestablished methods and at the same timeinclude an awaited instrumental innovation.The idea is to first detect a great number ofnatural targets as e.g. large rocks or treetrunks using terrestrial laser scanning (TLS) andthen to localize these with a terrestrial posi-tioning system (TPS). First the target objectswill be aimed at in their last known positionand scanned using a reflectorless measuringprecision tacheometre (STEMPFHUBER & WUN-DERLICH, 2004). The resulting 3D-model of eachstructure is then compared to the original tem-plate from the TLS (WUNDERLICH et al. 2005)and statistically checked for changes in itsposition in space. Later the internal cameraof the tacheometre, which is usually used forthe automatic aim on reflectors, will supportthe reflectorless aim by conduction a contourcomparison. First achievements have beenmade with this method at the Chair for Geo-desy (TUM) using the example of the automat-ic recognition of church spires as geodeticorientation (WASMEIER, 2003, 2004). Momen-tarily the technique is limited by the internalcamera, which, due to its fixed focus, can onlypicture objects in distances above 500 msharply. However, one manufacturer has justdeveloped a tachymeter containing a camerawith adjustable focus (WALSER & BRAUNECKER

2003), so that it can be expected to bebrought onto the market soon. Then the thirdphase is entered in which a great advance inthe operating efficiency can be achieved. Forsuch an instrument a control system can bedeveloped, which will allow it to automatical-ly search a slope for fitting natural objects,measure them and repeatedly check them fora change in position. Therefore it can be

installed at virtually any location and find tar-gets itself, to then start with their periodicalsurvey. In a symbiosis with the other two sys-tems, TDR and GNSS, a nearly perfect intelli-gence unit is produced, which enables toestablish an early warning system and tobetter understand the mechanical processeswithin a landslide.

3.3. Global Navigation Satellite SystemThe primarily contribution of the Universitätder Bundeswehr München to the alpEWASproject will be an all weather proof Low-Cost-GNSS-System which continuously re-cords movements on the surface of theinstable slope with the quality of a few mil-limetres. It concerns the development of aprototype of a GNSS-Monitoring Componentwhich in particular features the followingspecial characteristics:– Low-Cost-GNSS-Sensor technology (resp.

Low-Cost-GPS-Sensor technology);– WLAN-Communication between the sta-

tions (wireless data transmission);– autonomous power supply of the Rovers;– flexibility of the analysis through (at any time

adaptable) options of post processing;– possibility for remote maintenance and

remote inquiry of the system (remote desk-top operation);

– separation between data recording andessential evaluation (especially time seriesanalysis); defined ASCII-interface (»dualsystem«);

– possibility of incorporation of existing(proofed, powerful) program systems;

– open system to integrate other sensors resp.to be adapted to other sensors (»multisen-sorsystem«, component of a Geo-SensorNetwork).

After the installation the (entire) system – theGNSS-Monitoring Component is only a part ofthe planned Geo-Sensor Network – will con-tinuously provide measuring data. Its presen-tation and analysis are essential parts of thelater project phases (see below). Beside thedevelopment and trial of the three technologi-cal monitoring systems TDR, GNSS and TPS,their integration into the Geo-Sensor Network

108

is the centre of the scientific project. Thisincludes at first a purely technical and later asoftware-related, data base oriented combina-tion of the measuring results. This representsanother focal point of the work done by theUniversität der Bundeswehr München.Up to date the GNSS-Monitoring Componentde facto is a purely GPS-Component. If (assoon as) available, it should be changed toreceivers, which are able to register GALILEOand optionally GLONASS signals as well. Thisis presently not possible, but is, however,imminent. The planned set-up of the onlineLow-Cost-GPS-Systems (L1-receiver, no RTKavailability) provides four stations/receivers.Two respectively three stations of them areinstalled in range of the slide slope and thefourth receiver on a central station in the pre-sumably stable range in the near vicinity.The sketched configuration is currently testedin the scope of a still ongoing dissertation atthe Universität der Bundeswehr München. Be-side the OEM Boards Novatel Smart Antenna(receiver/antenna combination) different sen-

sors are also integrated in this test configura-tion like temperature sensors, voltmeters forcontrol of the battery voltage in the solarmode, geotechnical sensors such as extenso-meter and hydrostatic measuring devices. Atpresent a year-round field trial is coming up,the previous test structure is operating since afew months on the campus of the University ofthe Bundeswehr Munich. The system can beremotely maintained and controlled in theviewer modus (at any time) by thirds as well.At the computer centre (host system), wheretwo computers are provided for the systemcontrol, the access to 220 V power supply andto the internet (telephone) is necessary. Theother stations of the GNSS-Monitoring Com-ponent (rover stations) maintain a self-suffi-cient, autonomous power supply. The concep-tion for the Aggenalm landslide provides sole-ly solar panels and abandons an additional/alternative use of wind engines. This is guar-anteed by the sufficient dimensioning of thepanels and the buffer batteries. Demands tothe GNSS-Rover Stations are:

Figure 4: Measuring configuration of the GNSS-Monitoring Component at the test site Aggenalm

109

– preferably little signal shading of the orbitresp. space segment (no signal obstacles);

– quasi-optical view to the computer centre(with WLAN Access Point); obtainablethrough a raised set-up of the WLAN anten-nas (transmitting poles), the choice of theantennas and/or a repeater station (plannedin the financial means including necessarythe self-sufficient power supply, see below);

– proximity to TDR-Measuring Arrays, to beable to analyse the obtained informationoriented in the superior reference frame andintegrate this data into the Geo-SensorNetwork, as well as, the data of the TPS.

Technically speaking, the GNSS-MonitoringComponent should revert to Low-Cost-receivers with the possibility of a post pro-cessing. On the one hand for cost reasons(RTK-systems are 10-fold more expensive) andon the other hand for the advantages whicharise when using the option controlled analy-sis (comp. HARTINGER, 2001). Among others theupdate of the time interval is arbitrary, it usu-ally lies in the array of approx. 15 min. up to1 h. Beside other aspects this is also of impor-tance to the achievable accuracy of the baselines: the longer the time period in whichphase measurements in the OEM board areacquired, the better the positioning accuracymay be expected. The studies of the GPS-Sen-sor Novatel Smart Antenna, conducted at theUniversität der Bundeswehr München, show,that an accuracy in the scope of a few mil-limetres for the base lines can be achievedusing 15 min. intervals.

4. Integration and AutomationIn addition to the development and testing ofthe three measuring systems, their integrationinto a combined Geo-Sensor-Network is amain task of the research project (figure 5). Onthe one hand this applies to the developmentof techniques and software for the databaseoriented combination of the data, on theother hand also to the gain of scientific knowl-edge by means of the analysis and interpreta-tion of the data.WLAN (54 Mbps data rate) is earmarked forthe data transmission between the rover sta-

tions, at which the GNSS, TPS, TDR, piezome-ter and climate measurements are collected,and the computer centre. With WLAN clear(»quasi-optical«) visibility up to 3 km can bebridged; adequate antennas presumed. Withthe distinct slope morphology at the Agge-nalm landslide the direct visibility is severelylimited, so that a repeater station is planned.With regard to the development of a flexiblemonitoring system the incorporation of sucha relay station is also of fundamental interest,however. Compared to the usual transmissionpath (esp. with RTK-systems) WLAN offersseveral advantages and starts to be estab-lished. Basically every sensor with a RS232interface can be integrated to the net usinga COM server.Within the project the batch controlled pro-cessing will be carried out with the multiven-dor capability software GravNav Waypoint (seewww.waypnt.com). GravNav is thereby inte-grated in a LabVIEW-Programing, NationalInstruments (see www.ni.com/labview), whichmanages the sampling and selecting of thesensor knots. As a result of the LabVIEW/Grav-Nav-pre-processing the base lines (this meanstime information, 3D coordinates differences,assigned covariance matrices) are appliedthrough an open interface to the automatedevaluation system GOCA (GPS Control andAlarmsystem; developed at the Fach-hochschule Karlsruhe, see e.g. KÄLBER; et al.,2001a, b, see www.goca.info). GOCA pre-sents in this conception basically a highlysophisticated analysis tool for monitoringtasks, in which all features of the time seriestheory (filter operations as moving averageand spline approximation, trend estimation,Kalman-filtering etc.) are realized, as wellas, e.g. alarm functions (via SMS, e-mail,etc.) by the exceedance of thresholds. Fur-thermore, GOCA offers a nearly perfectedvisualization component. Also results fromtacheometric observations can be integrated(DANESCU, 2006).Finally however, GOCA may be replaced byany other program which enables the analysisof time series. For the combination with all theother data sources (TDR, TPS, meteo-sensors,

110

…) – this is not possible with GOCA – thedevelopment of an adequate, data base ori-entated and with GIS-functionalities equippedsoftware within this project is planned. Thisshould be done in a similar way as the pro-grams DIVA by the diploma thesis CRAUSE &ELING (2001) and the program ANALYSIS by

the diploma thesis GLÄSER & KNAUER (1999).Furthermore a cooperation and coordinationwith the participants of the joint project EGIFF(»Entwicklung geeigneter Informationssystemefür Frühwarnsysteme«) is arranged.The software component to be developedshould offer possibilities as an information sys-

Figure 5: Schematic illustration of an integrative 3D early warning system for instable slopes. GNSS and TDR continu-ously gather highly precise punctiform information about the movements on the surface and underground, respectively.The TPS in form of a reflectorless robot-tacheometre can gain displacement information covering most of the landslidearea. Furthermore triggering mechanisms are observed using piezometers (hydrostatic pore pressure) and a meteorolog-ical station (precipitation). When combining all the different measurements, a 3D model of the slope and its move-ments can be derived in temporal correlation to the triggering parameters. The measured data is transmitted to themaster station via WLAN and stored in a central computer. The data can be accessed and downloaded from a remotecomputer for further analysis and interpretation.

111

tem for the user on manifold ways of (inter-actively) getting access to the data. Here it isstated that the data from different compo-nents TDR, GNSS and TPS (»reaction quanti-ties«) are already pre-processed and if neces-sary reduced. In an ideal case the new soft-ware tool only will handle co-ordinates in acommon reference frame with respect to thereaction quantities.(Minimum-)Characteristics of the new soft-ware component to be developed are:– Input of pre-processed data from TDR, GNSS

and TPS; gauge data and meteo-sensors,storage in a data base, preferably MSACCESS;

– sophisticated possibilities to choose interest-ing sensors (Visual Basic- and/or MatLab-programming, to see the allocation of thesensors at the slide, activation via »click«);

– Description of single time series numericallyand graphically, including a zoom-function;

– Combination of any selectable sensor (e.g.comparison of impact quantities and reac-tion, comparison of adjacent sensors);

– alert function after reaching limiting values(thresholds);

– additional information for the sensors (digi-tal images, specifications, maintenance,remarks, …).

The software should be easy and intuitive tohandle, especially for third persons, who arenot involved to the project permanently, toget a quick overview. A modular composi-tion is intended. At the end only the mainuser interface must be changed to make thesoftware available for another project in asimilar way.In the course of the geologic examination ofthe received data first simple causal and tem-poral coherences between the parameters,which influence the slope (e.g. precipitation),and the occurring movements will be identi-fied. This easy analysis can be done within arelatively short period of time after starting themeasurements and will lead to the definitionof first threshold values for the early warningsystem. Then, through creation of numericalmodels of the instable slope using the ItascaUDEC and FLAC (Itasca) code, the geome-

chanical behaviour of the slope can beanalysed more precisely. The continuously col-lected measuring data is integrated into themodels leading to their constant improvement.The resulting better understanding of the trig-ger, mechanics and damage potential of thelandslide is used to optimise the definedthreshold values.

LiteratureCRAUSE, C. & ELING, D. (2001): Entwicklungeines Datenvisualisierungs- und Analysepro-gramms für die Überwachungsmessungen derOkertalsperre im Harz. – Diplomarbeit Univer-sität Hannover, Geodätisches Institut.

DANESCU, A. (2006): Case Studies Related toa New Approach of the GPS-Based OnlineControl and Alarm System Package Softwarefor Standard and non-GPS Application. –Diplomarbeit Universität der BundeswehrMünchen, Institut für Geodäsie.

GLÄSER, A. & KNAUER, C. (1999): Beurteilungvon Deformationsprozessen mit der Hilfe derFilterung von Zeitreihen und der Ableitung vonKonfidenzbändern. – Diplomarbeit UniversitätHannover, Geodätisches Institut.

HARTINGER, H. (2001): Development of aContinuous Deformation Monitoring Systemusind GPS. – Ingenieurgeodäsie – TU Graz (Hrsg.F.K. Brunner), Shaker-Verlag, Aachen 2001.

KANE, W.F., BECK, T.J. & HUGHES, J.J. (2001):Application of Time Domain Reflectometry toLandslide and Slope Monitoring. – TDR 2001– Second International Symposium and Work-shop on Time Domain Reflectometry for Inno-vative Geotechnical Applications, Infrastruc-ture Technology Institute, Northwestern Uni-versity, Evanston, Illinois, U.S.A.

O’CONNOR, K.M. & DOWDING, CH.H. (1999):GeoMeasurements by Pulsing TDR Cables andProbes. – CRC Press, Boca Raton.

SCHEIKL, M, GRAFINGER, H. & POSCHER, G.(2001): Entwicklung und Einsatz eines

112

automatischen Fernüberwachungssystemsbasierend auf einem Laserscanner. – in:Tagungsband zur 11. Int. GeodätischenWoche Obergurgl, Wichmann-Hüthig Ver-lag, Heidelberg.

SINGER, J. & THURO, K. (2006): Developmentof a continuous 3D-monitoring system forunstable slopes using Time Domain Reflec-tometry. – 10th IAEG Congress, Nottingham,United Kingdom, 6–10 September 2006, Geo-logical Society of London (in press).

SINGER, J., THURO, K. & SAMBETH, U. (2006):Entwicklung eines kontinuierlichen 3D Über-wachungssystems für instabile Hänge mittelsTime Domain Reflectometry (TDR). – Felsbau24 (3), 16–23.

STEMPFHUBER, W. & WUNDERLICH, TH.(2004): Leica System 1200: Auf dem Weg zurSensorsynchronisation von GPS und TPS fürkinematische Messaufgaben. – AVN, 111. Jg.,Heft 5, S.175–184, Wichmann-Hüthig Ver-lag, Heidelberg.

TOPP, G.C., DAVIS, J.L. & ANNAN, A.P (1980):Electromagnetic Determination of Soil WaterContent: Measurement in Coaxial Transmis-sion Lines. – Water Resources Research 16,Nr. 3, S. 574–582.

WALSER, B. & BRAUNECKER, B. (2003):Automation of Surveying Systems throughIntegration of Image Analysis Methods. – in:Proc. of Optical 3D Measurement TechniquesVI-1, Zürich.

WASMEIER, P. (2003): The Potential of Objectrecognition Using a Servo-TacheoemterTCA2003. – in: Proc. of Optical 3D Measure-ment Techniques VI-2, Zürich.

WASMEIER, P. (2004): Potenzial der Objekt-erkennung mit dem ServotachymeterTCA2003. – Geomatik Schweiz, Heft 2.

Wunderlich, TH., SCHÄFER, TH. & ZILCH, K.(2005): Schadenserkennung an einer Spann-betonbrücke durch reflektorlose Deformation-smessungen. – in: Festschrift 60. Geb. Prof.Bancila, S. 60–68, TU Politehnica Timisvar.

113

AbstractAt present the analysis and preparation ofinformation are particularly critical points ofan early warning chain. The responsible deci-sion makers are usually confronted with hugeamounts of structured and unstructureddata. To enable reliable early warning, theavailable data must be pre-selected, analysedand prepared. The decision makers should beprovided with a reliable and manageableamount of information for the warning deci-sion and for taking preventive measures. Inthe introduced joint project »Development ofsuitable information systems for early warn-ing systems« (EGIFF) components of an infor-mation system for early recognition of geo-logical hazards will be developed and exam-ined in a mass movement scenario. In partic-ular the improvement of the informationanalysis and preparation is researched. There-fore techniques like GIS, numerical simula-tions, Spatial Data Mining, geodatabases andthe application of linguistic methods will becombined. A particular focus will be put oninnovative methodical investigations, whichwill serve to merge new technologies into theoperational workflow of early warning and to

establish an efficient and reliable medium fordistributing information to the responsiblehazard/crisis managers.

1. Overall Objectives of the ProjectThe central components of an early warningsystem are the recognition of the threats, theassessment and evaluation of the danger, thedissemination and communication of thewarning, as well as the public reaction to thewarning (Smith, 2004). The effectiveness of anearly warning system largely depends on thetransformation of the event recognition intothe report of warning to the population (Dikauand Weichselgartner, 2005). Obviously theanalysis and information preparation are par-ticularly critical points of the early warningchain. They provide a basis for the warningdecision and the risk estimate for the extent ofthe natural event.The objective of the joint project EGIFF is toimprove the early warning chain through theconception and development of appropriatecomponents for early warning systems. Here-by the analysis and preparation of informationare the basis of the warning and the risk esti-mation of the forthcoming geological natural

Development of suitable information systemsfor early warning systems (EGIFF)

Breunig M. (1), Reinhardt W. (2), Ortlieb E. (2), Mäs S. (2), Boley C. (3), Trauner F. X. (3), Wiesel J. (4),

Richter D. (4), Abecker A. (5), Gallus D. (5), Kazakos W. (6)

(1) Institute for Geoinformatics and Remote Sensing (IGF), University of Osnabrück, Kolpingstr. 7, D-49069 Osnabrück,

Email: [email protected]

(2) Geoinformatics Working Group (AGIS), University of the Bundeswehr Munich, Heisenberg Weg 39,

D-85577 Neubiberg, Email: [email protected]

(3) Institute for Soil Mechanics and Foundation Engineering, University of the Bundeswehr Munich,

Heisenberg Weg 39, D-85577 Neubiberg, Email: [email protected]

(4) Institute of Photogrammetry and Remote Sensing (IPF), University of Karlsruhe, Englerstr. 7, D-76128 Karlsruhe,

Email: [email protected]

(5) Research Centre for Information Technologies (FZI) at University of Karlsruhe, Haid-und-Neu-Str. 10–14,

D-76131 Karlsruhe, Email: [email protected]

(6) disy Informationssysteme GmbH, Stephanienstr. 30, D-73133 Karlsruhe, Email: [email protected]

114

event. The main research objectives of the pro-ject include:1. Conception of a distributed component

architecture of an information system forearly warning systems;

2. Geotechnical evaluation of mass move-ments with the help of available events;

3. Coupling of numerical simulations and GIS;4. Modelling and visualization of spatial

relationships and their uncertainty as wellas their extraction from textual andnumerical data;

5. Transfer of Data Mining and analysismethods to spatially referenced data(Spatial Data Mining), as well as processingof structured and unstructured data (freetexts of disaster and damage messages);

6. Web-based alerting;7. Geodatabase support for handling data of

different scenarios as decision basis for thegeotechnical evaluation of mass movements;

8. Development of concepts and methods forthe support of the evaluation of risks and theirprototypically implementation and finally

9. Evaluation of the developed prototype onthe basis of concrete geological data andapplication scenarios.

2. Application AreaA number of places in the German alpineupland and the Bavarian Alps seemed to besuitable investigation areas for exemplary land-slide simulations. Main selection criteria werethe availability of detailed geographical dataand sensor measurements for a longer periodof time, a coherent and comprehensive geolo-gy to verify the gained methodology in a gen-eralized way and a potential risk. After inves-tigations of different landslide areas a part ofthe Isar valley in the south of Munich, next toPullach and Neugrünwald (see Figure 1), hasbeen selected for the further studies.

Figure 1: Application Area »Isarhänge«

115

Figure 2: Slopes in the Application Area »Isarhänge«

Figure 3: Overall Architecture

In this area, the height difference of the slopereaches up to 40 meters and the potentiallyendangered human infrastructure is locateddirectly at the edge of the slope. Because ofthe high risk potential, the area is observed by

the responsible authorities (Bayerisches Lan-desamt für Umwelt), and corresponding incli-nometer and extensometer measurements andgeodetic surveys are available.

116

3. Description of the subprojectsThe project is subdivided into three subpro-jects. Figure 3 illustrates these subprojects andtheir interaction. The following subchaptersprovide an insight into the objectives andapplied methodologies of the subprojects.

3.1. Subproject I »Developmentof an interconnected informationand simulation system«

Responsible: Conrad Boley, Institute for SoilMechanics and Foundation Engineering (IBGN)and Wolfgang Reinhardt, GeoinformaticsWorking Group (AGIS) both University of theBundeswehr Munich

3.1.1. ObjectivesThe main aim of this subproject is the concep-tion, prototypically implementation and evalua-tion of an interlinked three dimensional Numer-ical Simulation System (SIMS) and a Geograph-ical Information System (GIS). Its research con-tributions are, in particular, the development ofcoupled information and simulation models formass movements with a user-friendly control ofthe complete system. Another focus is put ona more precise predictability and a more exactdetermination of exposure of slopes. Thereforea transferable numerical simulation algorithmfor landslide movements will be realised withina simulation system. Thus, it is possible, basedon existing field data (sensor- and GIS data, forexample DTM) and virtual exogenous scenarios,to assess the slope stability, future systembehaviour and potential risk scenario. As aresult, a »communication« between the sensormeasurements, GIS and the geotechnicalmodel is enabled.The interconnection between SIMS and GIS isschematically shown in figure 4. The processstarts with the selection of relevant parameters(Input Data) which may have influence on theoccurrence of landslides. These parametersinclude basically temperature, precipitationand slope geometry. The parameter transfer iscontrolled by the GIS. In the simulation sys-tem, the slope stability and deformation is cal-culated and, if possible, used for prediction of

the future deformation behaviour. Results ofthe simulation are parameters such as hazardpotential, deformation and movement vectorsand stability indices. These results are trans-ferred to the GIS for visualization and, as faras possible, enrichment with other informa-tion. Furthermore, the data can be checkedagainst rules to support the decision makingprocess of the user.The developed concepts for the interconnec-tion of the GIS and the simulation system willbe prototypically implemented and evaluated.With this prototype of a Decision Support Sys-tem (DSS), it can be demonstrated how thelinking of numerical simulations and GIS canimprove predictions of landslide hazards andcalculation of vulnerability, and how theuncertainty of and lacks within the data canbe considered. Additionally, it will be verifiedhow the output of simulations, enriched withGIS analytic methods, can optimize the deci-sion making process for the prevention of ca-tastrophic hazards.Within the overall system architecture (see Fig-ure 3), this DSS is integrated with the geo-database (Subproject III) and the data miningsystem components (subproject II). The geo-database provides the input data for the simu-lation and the modelling of slope stability andstores the simulation results. The event mes-sages detected by the data mining system willalso be incorporated into the decision support.

3.1.2. Concepts and MethodologyIn the following the main work packages ofsubproject I are described:

1. Concept of an overall architectureIn cooperation with the project partner Uni-versity of Osnabrück, a suitable architecture ofthe overall system consisting of SIMS, GIS, geo-database, in-field installed sensors as well asothercomponentswillbedeveloped.Thearchi-tecture should enable a seamless communi-cationbetween individual system components.

2. Data integration and modellingThe data models required for the numericalsimulation and GIS will be designed. The

117

main focus of the modelling process lies inthe combination of stratigraphical, litholog-ical and geotechnical data with geometricsensor data, long term physical series ofmeasurements, as well as the requirementsof visualisation and further processing ofthe simulation results. Subsequent to mod-elling, the available data of the test areawill be adopted and integrated.

3. Conception of geotechnical modelsand their interpretationNatural slopes often consist of several dif-ferent soil types, whose mechanical prop-erties have to be described with differentconstitutive equations. For a realistic char-acterisation of the deformation behaviourof different soil types, the validity of dis-tinct constitutive equations for each rele-vant material has to be analyzed andassessed. Therewith the application ofappropriate constitutive equations for dif-

ferent soil types will be enabled. To applythe best suitable constitutive equations, therequired material parameters have to beidentified and a concept for the determi-nation of these parameters developed. Thematerial parameters will be stored in adatabase, so that the final numeric modelcan access the required parameters.For numeric modelling of landslides initial-and boundary conditions have to bedefined. These include the geometry of theslope as well as superimposed load andmeteorological influences, etc. Due to thecomplexity of the actual initial- and bound-ary conditions, a simplification of thenumerical model is inevitable.Numeric modelling demonstrates not onlythe deformation of the slope but also theslope stability. For the determination of theslope stability failure criteria have to bedeveloped, which may vary for differentmaterials and different constitutive equa-

Figure 4: Interconnection between simulation system and GIS

118

tions respectively. With the definition ofthese criteria the material properties, thestate of stress and strain and the desiredsecurity level have to be considered, as wellas the strain rate and stress level. Based ona time variant 4D-simulation, calibratedwith backwards analysis for past events,probabilities for future scenarios can be cal-culated. By comparing predicted deforma-tions with measured displacements, themodel can be adjusted and improved.

4. Conception of the Decision SupportSystemThe criteria for determining the stability ofslopes enable the mapping of landslideprone areas and vulnerable zones. Thismeans the slopes can be divided intohomogenous areas according to the degreeof actual or potential hazard. Furthermore,methods will be developed for refining andlinking the simulation results with otherinformation (e.g. from subproject II) and forvisualizing the simulation events. In addi-tion, the simulation results can be linkedwith decision rules in order to support andguide the user. For example, dangerous sit-uations and adverse constellations thatmay be predefined and formalized in thesystem can be automatically revealed to theuser as advice. The main goal is to estab-lish an efficient and reliable medium fordistributing information to the responsiblehazard managers. Furthermore, it will beinvestigated if and how uncertainties of thedata used in the simulation and subse-quent processes can be modelled. In par-ticular, for visualization and for the supportof the user in the decision-making processthese uncertainties should be recognizable,in order to allow for validation of theresults by the user. Moreover it will beinvestigated if and how this validation ofuncertainties can be done automatically.Another emphasis is put on the convenientpresentation of the extensive simulationresults to the user and the development ofa user friendly and intuitive system control.

5. Definition of the interfaces between GIS/geodatabase and the simulation componentThe main interfaces that need to bedefined are for the transfer of the inputdata and the simulation results betweenthe GIS and the simulation component,and for the connection of the GIS to thegeodatabase. Therefore available interna-tional standardized interfaces will be exam-ined and, if necessary, restricted, extendedor adjusted. For the bidirectional access tothe geodatabase, standardized interfacesof the OGC (Open Geospatial Consortium)will be taken into consideration. For thelinking between the simulation system andthe GIS also standards like HLA (High LevelArchitecture) will be analysed.

6. Prototype developmentThe developed concepts will be prototypi-cally implemented and verified with data ofthe above mentioned application area.Practical tests will be carried out to deter-mine the system feasibility, performanceand reliability. Furthermore a transferabilitystudy shall analyse the applicability of theconcepts and solutions to areas with simi-lar and also varying geologic and boundaryconditions and data availability.

3.2. Subproject II »Integration, analysisand evaluation of vague textual descriptionsof geoscientific phenomena for early warningsystem support«

Responsible: Peter Lockemann, FZI Karlsruhe;Joachim Wiesel, Institute of Photogrammetryand Remote Sensing (IPF), University ofKarlsruhe; Wassilios Kazakos, disyInformationssysteme GmbH, Karlsruhe

3.2.1. ObjectivesThe goal of subproject II is the collection, pre-processing and analysis of relevant structuredand unstructured data (numerical and textualmeasurements, observations and descriptionsof specialists and laymen (citizens)) related tonatural hazards in form of disastrous massmovements, with a focus on development and

119

application of novel computational methodsaimed at a combination of results of analysesfrom heterogenic data sources, followed by aprototypical implementation as suitable com-ponents of an early warning system.For this goal, the following central tasks havebeen identified:– development and application of techni-

ques and methods aimed at an automaticextraction of early warning-relevant in-formation and spatial references from tex-tual messages,

– transfer of suitable techniques andmethods for automated data analyses andan application thereof in the domain ofapplied geosciences,

– development and application of novelmethods aimed at a combination of re-sults of analyses of structured andunstructur-ed data related to naturalhazard phenomena of the type »disastrousmass movement«

– design of ergonomic user interfaces for thetask of complex data analyses in context ofearly warning systems, with a focus on userswith none or no profound IT knowledge

Besides concentrating on disastrous massmovements as the context for developmentand application of appropriate techniques andmethods, a high degree of transferability toother natural hazard types or even analyticaltasks in the domain of applied geosciences isregarded as desirable. The latter is also seen aspart of a general objective to maximize scien-tific connectivity as well as a prospective com-mercial value of the project’s results.

3.2.2. Concepts and MethodologyIn the following the three main work packagesof subproject II are described. They will beworked on collaboratively by FZI, IPF and disyInformationssysteme GmbH, Karlsruhe.

1. Conceptual workIn the first year of the project, the moregeology-oriented and the more computerscience-oriented project partners will cometo a common understanding of the pro-ject’s goals, and develop well-defined,

effective and efficient ways of collabora-tion. Based on a deeper understanding ofthe application domain with its specificrequirements, characteristics and potentialproblems, and in close cooperation withthe other subprojects, a comprehensive sys-tem vision and architecture will be devel-oped, including a set of accepted interfacesand – where necessary – shared data struc-tures. On this basis, example-driven tech-niques and algorithms will be specified tobe implemented later (year 2). In detail, weaim to focus on the following tasks:

1.1 Analysis of the application domainIn a first step of the analysis of the appli-cation domain, a glossary of central con-cepts and terms for the subproject part-ners will have to be developed in order tofacilitate communication with other sub-projects and with end users. At the sametime, relevant functional and non-func-tional requirements from end users’ per-spective and constraints for the systemsto be developed have to be identified.

1.2 Investigation of relevant prior workIn this task, focus is set on an investiga-tion of relevant prior work within andoutside the domain of applied geo-sciences and an analysis with respect toan applicability and usefulness in thegiven problem setting.A first part of the work will include areview of SOKRATES, a system developedby FGAN/FKIE in Wachtberg (Schade andFrey, 2004). Relying on a combination ofmethods from natural language process-ing and techniques of knowledge repre-sentation, the system can be used to dis-play events on a tactical map by means oflimited automated interpretation of battlefield reports. The combination of tech-niques employed by SOKRATES can serveas a methodological basis for parts of thesystem envisioned in subproject II. Due todifferences in application context, a num-ber of adjustments will have to be made,whereas parts of the system may be

120

replaced by Semantic Web de facto stan-dards or Open Source developments. Aparallel thread of work will focus on aninvestigation of data mining techniqueswhich could be transferred into thedomain of applied geosciences in order todevise, develop and select suitable tech-niques, algorithms and tools (Lindner,2005). In this context, assessment of algo-rithms and tools will also comprise anevaluation of the transferability/applicabil-ity of selected representation formalisms,methods and systems to other real-worldapplications, as well as their interoperabil-ity with established tools.

1.3 Collection and detailed analysisof input dataThe next step in task chain will concernthe collection and detailed analysis ofavailable data for data mining algorithmsto be developed. Available structured andunstructured data (text) will have to becollected, formally described and storedfor application and test. Additionally, fur-ther work will address the issue of acqui-sition of expert knowledge needed for theinterpretation of information as the basisfor planning actions to avert damages ofpeople and property.

1.4 Design of interfaces and specificationof requirements for shared datastructures and system componentsIn this task, focus is set on the design ofinterfaces and requirements for shared datastructures and system modules in order toensure interoperability with other subpro-jects, in particular with respect to specificrequirements introduced with computation-al representation of three-dimensional spa-tial data and the representation of vague-ness and uncertainty resulting from ananalysis of textual spatial references.

1.5 Specification of systemsto be implemented (end user perspective)The next steps, subsumed in subtask 1.5,will deal with a specification of systems

and methods in the form of use cases,including a coordination of use cases withthe other subprojects’ use cases and anevaluation of the use cases by end users.Another task will be concerned with a gen-eral assessment of transferability, scientificconnectivity and a commercial exploitabili-ty of the systems to be developed, withrespect to different applications in the con-text of disaster management (early warn-ing systems) or data analysis in the domainof applied geosciences.

1.6 Specification of algorithms to beimplementedIn this task, three parallel threads of workwill focus on the following subtasks:In a first subtask, necessary changes andmodifications to the SOKRATES systemwith respect to the given setting will bedefined. For the task of the analysis ofstructured spatial data, suitable data min-ing techniques will be specified. Addition-ally, the task of a rough specification of agraphical user interface of the system willbe addressed.The tasks in the first year of the project(conceptual work) will be worked on col-laboratively by all partners.

2. Implementation-technical realizationIn the second year, the focus of subprojectII will be set on an implementation of thealgorithms previously specified and thenecessary system adjustments. Dependingon the progress and state of work in othersubprojects, further functional aspects willbe defined which will be implementedlater. It is seen as reasonable to delay thespecification of these functional aspectsinto this phase of the project until a condi-tion is met assuming basic work in theother subprojects is completed and a firstfeedback concerning the results of imple-mentation etc. is received. In particular, thefollowing packages will be processed:

121

2.1 Implementation of text analysis methodsIn this work package, previously specifiedadjustments, modifications, and enhance-ments to the SOKRATES system will beprototypically implemented. Subsequently,the system will be tested, making use ofthe data collected.

2.2 Implementation of spatial data mining/suitable methods for the analysisof structured spatial dataIn this work package, the techniques andmethods previously specified will beimplemented, adapted to, and integratedinto the project’s common softwareframework. Adaptations of existing tech-niques may also include handling of threedimensional spatial data, methods for effi-cient processing and techniques for pro-cessing data characterized by inherentvagueness or uncertainty.

2.3) Specifying mixed analysis methodsand the integrated user interfaceRelying on results from 2.1 and 2.2, spec-ifications of methods to be implementedin the third year of the project will bedeveloped. A first part of the work willconsist of devising methods for combinedprocessing of structured data and post-processed unstructured data (i.e. resultsof analyses of textual messages). In thistask, central focus will be set on a moredetailed investigation of the concept of»spatial coincidence« and certain aspectsof methods for combined processing ofheterogenic data, in particular related to aquantitative treatment of results under acommon decision-theoretic framework.Secondly, an integrated user interface willbe specified which will allow non-IT-expertsto efficiently make use of the multitude ofdata sources and algorithms. Here, concep-tual work related to problems of suitablerepresentation of spatial data will be putinto practice, including, but not limited to,a pragmatic choice of tools for formulatingand testing hypotheses or for queryingcomplex data and knowledge bases.

In context of visualization of spatial data,characteristics of the data related to relia-bility and precision of spatial referenceswill have to be presented in a way satisfy-ing requirements for adequate usability. Inparticular, previously established qualitieslike priority or urgency of information andtheir possible interpretations as well asscale-dependencies will be major aspects.

3. EvaluationIn the third project year, two major threadsof work will be pursued. In the first sub-task, implemented algorithms and systemswill be documented, tested, evaluated withreal-world data and validated against dif-ferent defined use cases. This will be donein coordination with other subprojects, andin touch with international scientific com-munity. In the second subtask, a parallelthread of work will be started, focussing onan implementation of combined analysisalgorithms for structured and unstructureddata. Additional work will include animplementation of ergonomic user inter-faces, test-driven removal of errors anddeficiencies based on results of parallelevaluation work, and finally an integrationof the implementation with the other sub-projects’ results.

3.3. Subproject III »Geodatabase supportfor the geotechnical evaluationof mass movements«

Responsible: Martin Breunig (Consortiumcoordinator), Institute for Geoinformatics andRemote Sensing (IGF), University of Osnabrück

3.3.1. ObjectivesIn this subproject, the primary geological data,provided by the Bayerisches Landesamt fürUmwelt, will be modelled and managed in ageodatabase management system. The man-agement of the geological spatial and time-related primary data enables its usage foranalysis, simulation, and 3D visualization atany time. In a second step, finite elementmodels and their results will be stored in the

122

geodatabase. This supports the geotechnicalevaluation of mass movements through theassistance of suitable database queries to bedeveloped during the project. Finally, thedevelopment and implementation of geomet-ric/topological and time-related operations asinput for spatial data mining methods shallprovide new spatio-temporal patterns for therecognition of mass movement hazards. Thedeveloped methods will be evaluated on thebasis of concrete geological application datafrom the selected area of investigation.

3.3.2. Concepts and MethodologyGeodatabases are used for archiving and forproviding fast and efficient access to geodata.This enables for the reuse of geodata, e.g. forupcoming new natural events. In addition, ageodatabase can also manage models andtheir results and even become active, e.g. bycomputing geometric intersection queriesbetween a set of geoobjects.The subproject pursues three objectives:1. Geometric and topological management

and 3D visualization of existing geologicalprimary data and time-dependent data.

2. Management of FE models and their resultsfor the use in geotechnical evaluations ofmass movements.

3. Development and implementation ofgeometric/topological and time-depen-dent operations as input for spatial datamining methods.

For the first objective, selected test data fromthe investigation area (see section ApplicationArea) will be examined and managed in a geo-database in close cooperation with the geo-technical group Boley (subproject I). Thisincludes a first interpretation of the primarydata. In particular the representation of dis-continuities (e.g. with already existing faults inthe slope) will be taken into account. This goalcan use the developments implemented in(Breunig et al., 2005) and benefit from thisdevelopment of geodata management forcomplex geological models (surface layer andvolume models). The available data structuresand operations have to be examined for thesuitability of the special requirements by mass

movements. Where necessary these need tobe adapted and extended to fit the new spe-cial requirements.The second goal basically concerns the man-agement of the input data and results from 3Dmodel calculations within the geodatabase.This will be done in close cooperation with thegeotechnical group of Boley. With the FEmodel calculation a second kind of data inter-pretation will be realized. The management ina geodatabase creates new possibilities of fur-ther use of such modelling results and foroverlaying with further information like utiliza-tion data. For example the information »Howis the area under the endangered slopeused?« is relevant for early warning and there-fore needs to be coupled with the model cal-culation. Homogenized and/or converted geo-metric data models from the model calcula-tions and the geodatabases have to be exam-ined. Also the information computed for eachcell (scalars and vectors) of the FE model hasto be considered.Furthermore, the results of a larger number ofmodel calculations can be stored in the data-base and on later demand they can be com-pared and queried from different points ofview. For example, versions or scenarios ofmodel calculations can be stored and com-pared with each other. Additionally, taking theuse of a »4D model« into consideration, i.e.different time steps of the FE model, the datafor the temporal analysis of mass movementscan be managed in the geodatabase.It is also of importance to store error marginsand their accuracy in the database. Here werefer to subproject II. Intervals need to bestored in the database. Passing over or fallingshort of these intervals will cause an action ofthe database. From the database side, furthergeometric/topological constraints and/or in-tegrity conditions can be realized and includ-ed in the analysis of mass movement endan-germent. The thematic information connectedwith the geometry has to be extended in away that the model calculation receives infor-mation about the different semantics of geo-metrically equal geometries (e.g. fault surfacesversus boundary surfaces). In this context we

123

speak of »thematic colouring« of the model,which indicates to the user, for example, atwhich locations danger zones exist. After-wards the results are used in the data miningcomponent (subproject II).The third goal consists of developing geomet-ric/topological and time-dependent databaseoperations, which support the geotechnicalanalysis of mass movements. Of concern arenot only simple range calculations betweenpoint geometries (like it is the case in classical2D GIS buffer operations). Also the move-ment/speed of complex 3D geometries and thedirection of the movement have to be consid-ered. The implemented operations can be usedafterwards in the data mining methods, in orderto determine a priori unknown spatial patternsfrom the combination of spatial and non-spatialattributes for upcoming mass movements.Methodically, one of the principal purposes ofobject-oriented software technology, thereusability of individual components, will beused during the software development, inorder to react as flexible as possible to newrequirements of future natural events.

4. OutlookThe concrete research results of the joint pro-ject will be:– Geotechnical evaluation of mass movements;– Reusability of the geological and geotechni-

cal primary data and models relevant forearly warning through management in ageodatabase;

– Employment of spatial data mining methodsas support of the warning decision for natu-ral events;

– Development of new methods for the visualrepresentation and decision support of earlywarning relevant information from alphanu-meric data and texts including fuzzy andincomplete data;

– Prototypical implementation of the develo-ped methods and concepts and evaluationwith the application scenario »mass move-ments« (e.g. landslides);

– Improvement of the reliability of alarms forthe early warning of natural events/hazardsby employment of the developed methods.

The evaluation of the project results on thebasis of concrete application scenarios andtheir economic utilization by the partner com-pany are part of the project. The close coop-eration with the Bayerisches Landesamt fürUmwelt guarantees the direct application ofthe research results.

AcknowledgementsThe funding of the research project »Develop-ment of suitable information systems for earlywarning systems« (Entwicklung geeigneterInformationssysteme für Frühwarnsysteme) bythe German Ministry of Education and Research(BMBF) by grant no. 03G0645A et al. withinthe framework of the geotechnology initiative(http://www.geotechnologien.de) is gratefullyacknowledged. The responsibility for the con-tents of this publication is by the authors. Wealso thank the Bayerisches Landesamt fürUmwelt (LfU, www.lfu.bayern.de) and the Bay-erisches Landesamt für Vermessung und Geoin-formation (LVG, www.lvg.bayern.de), for pro-viding data for the application area.

ReferencesBreunig, M.; Bär, W.; Häussler, J.; Reinhardt,W.; Staub, G.; Wiesel, J. (2005): Advancementof Mobile Spatial Services for the Geosciences.In: Data Science Journal, International Councilfor Science (ICSU)

Dikau, R., Weichselgartner, J. (2005): Der un-ruhige Planet – Der Mensch und die Natur-gewalten, Wissenschaftliche Buchgesellschaft,Darmstadt, 191 S.

Lindner, G. (2005): Algorithmenauswahl imKDD-Prozess. Dissertation an der UniversitätKarlsruhe, Institut AIFB, 2005.

Schade, U.; Frey, M. (2004): Beyond informa-tion extraction: the role of ontology in militaryreport processing. In: Bachberger, E. (Ed.):Proc. of KONVENS 2004, Vienna, Austria.

Smith, K. (2004): Environmental Hazards:Assessing Risk and Reducing Disaster, London.

124

Exupéry: Managing Volcanic Unrest –The Volcano Fast Response System

AbstractVolcanic unrest and volcanic eruptions are oneof the major natural hazards next to earth-quakes, floods, and storms. In the frameworkof this project we will develop the core of amobile Volcano Fast Response System (VFRS)that can be quickly deployed in case of a vol-canic crisis or volcanic unrest. The core of thesystem builds on established volcanic monitor-ing techniques such as seismicity, grounddeformation, and remote sensing tools for gasmeasurements. A major novelty of this mobilesystems is the direct inclusion of satellite basedobservations to deduce ground deformation,to detect hazardous gas emissions and tomonitor thermal activity. The backbone of theVFRS is a wireless communication network tiedto a central data base that collects data fromthe instruments in the field. Using well definedexchange protocols that adhere to interna-tional standards the system will be open fornovel and promising tools like geoelectricsoundings and new spectroscopic methods tomonitor gas chemistry during volcanic unrest.The raw data collected by the VFRS will be fur-ther analyzed and fed into different models toconstrain the actual state of activity at the vol-cano and to set alert levels. Finally, all infor-mation and results will be made availablethrough GIS based visualization tools to com-municate the results to local decision makers.

1. Scientific reasoning

1.1. IntroductionA great majority of the world’s potentiallyactive volcanoes are unmonitored. Less than

twenty-five percent of the volcanoes that areknown to have had eruptions in historicaltimes are monitored at all, and, of these, onlyabout two dozen are thoroughly monitored(Ewert and Miller, 1995). Moreover, seventy-five percent of the largest explosive eruptionssince 1800 occurred at volcanoes that had noprevious historical eruptions (Simkin andSiebert, 1994). Being able to quickly react tovolcanic unrest at so far not well or unmoni-tored volcanoes is therefore a social challengenot in the least because the danger associat-ed with volcanoes is not only restricted to theireruption, but also includes earthquakes, dan-gerous gases (lake Nyos, Cameroon; Sigvalda-son, 1989), flank movement and other defor-mation, tsunamis (Stromboli, Italy; Bonaccorsoet al., 2003), landslides, and even climaticchanges (e.g. eruption of Mt. Pinatubo, 1991;McCormick et al., 1995). Defining criteria bywhich to forecast volcanic eruptions is there-fore the most fundamental goal of vol-canological research and it is a mandatory pre-requisite for any successful hazard mitigationstrategy associated with volcanic activity andcritically depends on a full understanding ofvolcanic systems.The basic question of why volcanoes becomerestless and erupt is not trivial. In fact, it is afundamental problem because only a smallfraction of magmas generated at depth everreaches the surface (Crisp, 1984). One of thekeys to understand the eruptive potential of avolcanic system is our ability to characterizethe actual state of stress of a volcanic system(e.g. Walter et al, 2005) and to understandhow susceptible the system is to small param-

Hort M. (1), Wassermann J. (2), Dahm T. (1) and the Exupéry working group

(1) Inst. für Geophysik, Universität Hamburg, 20146 Hamburg

(2) Department für Geo- und Umweltwissenschaften, Universität München, 80333 München

125

eter changes (e.g. pressure, temperature,water content etc. see also Hill et al, 2002).Whereas the first task involves proper moni-toring strategies including novel ground andspace based observation methods, the secondone is concerned with understanding theresponse of volcanic systems to various forcesacting on them. This requires an in depthmodeling approach to understand so far unex-plored internal feedback mechanisms in vol-canic systems. Only then we will be able toidentify a near critical state, and understandhow small changes in system parameters maywell be the decisive factor for the »final push«of magma ascent.Is there solid ground to believe that forecast-ing volcanic eruptions is a reasonable goal?Yes, because– volcanoes are basically point sources allow-

ing focused research strategies,– their number is limited and only a few volca-

noes are born each century,– volcanoes harbor an unparalleled event stra-

tigraphy and chronology, and– many different types of chemical and physi-

cal precursors can be recognized and moni-tored weeks or months prior to an eruption.

Classic success stories are the prediction of theeruptions of Mt. St. Helens, USA, (Lipman andMullineaux, 1981; Swanson et al., 1983) the1991 eruption of Mt. Pinatubo (Newhall andPunongbayan, 1996) and the crisis at Merapivolcano in 2006. In all cases thousands of liveswere saved because of timely evacuation.These predictions were only possible becauseof combined geodetic, seismologic, petrologic(in case of Pinatubo) and remote sensinganalysis. Especially the detection of differentprecursors points towards the importance of amultiparameter monitoring strategy that willultimately lead to a proper characterization ofthe volcanoes state of activity. However, thenature of most precursor signals of active vol-canoes has not been deciphered satisfactorilyand many volcanic eruptions still occur unex-pectedly. Successful volcano understandingand hazard assessment therefore necessitatessignificant advances in modeling processesinside volcanic systems.

1.2. Mechanisms know and suspectedto trigger volcanic eruptionsOver the course of the last two to three cen-turies several mechanisms triggering volcanicunrest have been identified. Regardless ofwhich mechanism one is looking at, the finalquestion is always:Does the state of stress inside the systemchange in response to changing system para-meters? Only if this is the case, then there is apotential for unrest and an eruption. There-fore, all mechanisms known to trigger volcanicunrest do change the state of stress of the sys-tem, some in a very direct way (e.g. injectionof magma into a confined system), some in amuch more subtle manner. A successful earlywarning strategy therefore critically dependson identifying and detecting parameters thatreflect the state of stress of volcanic systemson very different time scales. Well establishedtrigger mechanisms involve magma injection(e.g. Kilauea volcano, Hawaii, Eaton andMurata, 1960; Tilling and Dvorak, 1993) andmagma mixing (e.g. Pallister et al, 1992;Eichelberger, 1995), volatile exsolution (e.g.Holloway, 1976), and magma/water interac-tion (e.g. White and Houghton, 2000). Im-proved monitoring techniques and spacebased observations have led to the identifica-tion of several new trigger mechanisms likeseismic energy radiated by earthquakes (e.g.Barrientos, 1994; Marzocchi et al., 2002;Moran et al., 2002) tectonic stresses (e.g. Nos-tro et al. 1998), meteorological (e.g. Violetteet al., 2001; Hort et al., 2003; Richter et al.,2004) and climatic conditions (e.g. Schmincke,2004), gravitational loading (Borgia et al.,2000), and tidal forces (e.g. Emter, 1997; Tol-stoy et al., 2002).

2. Objectives and workplanMajor motivation for this research aside frombroadening our fundamental understanding ofvolcanic systems are the fast growing popula-tion, the increasing number of megacities, andthe complex communication, transport andsupply networks – all of which have resultedin a dramatically increased vulnerability ofmodern society especially near large volcanoes

126

(e.g., Naples, Catania, Tokyo, Mexico City,Seattle, Manila, Yogyakarta, Managua). Be-cause some of the high risk volcanic systems(especially in the EU) are already monitored invarious ways, we propose not to develop yetanother monitoring system but to develop thecore of a prototype of a mobile Volcano FastResponse System (VFRS) that can be deployedon a volcano in case of unrest or a crisis cer-tainly only upon specific request form the gov-ernment of the country the volcano is locatedin. The main idea behind this system is:– that it can be installed fast due to intelligent,

cable-free communication between the dif-ferent stations and a data center,

– that a larger number of stations can bedeployed,

– that all data are collected in a central database including the data from an existing net-work (open system) if desired by local scien-tists and authorities,

– that models are developed to derive activityparameters out of the recorded data,

– that the data are visualized and partially ana-lyzed in real time, and

– that objective and reliable data evaluationsare carried out including recommendationsfor crisis management.

The system is intended to densify an existingnetwork including some novel monitoringparameters that are regularly not observed butdo allow further insight into the processesinside the volcanic edifice. The VFRS includesa package of programs that uses the recordeddata to model and thereby better constrainthe complex internal feedback mechanisms inthe volcanic system to further improve earlywarning. All data are collected in a centraldatabase that holds all raw data. All resultsfrom specific data analysis and modeling arefed into a second database. A GIS based visu-alization system will display these metadata aswell as raw data in real time providing a con-tinuous view on the systems activity.In case of a deployment of the VFRS all infor-mation gathered by the VFRS will be discussedwith and provided to local experts and author-ities to aid their assessment of volcanic activi-ty. This is a very sensitive issue and requires a

great amount of diplomacy. It is clear that anyofficial statement regarding the state of activ-ity including setting alert levels or call evacua-tion can only be made by local authorities andnot by us. Our task will be to provide addi-tional information and parameters but we arenot in a position to officially communicatethose results nor can we call for any actions tobe taken.During the three years of funding we will cer-tainly not be able to deliver a perfect mobilevolcanic early warning system. However, wewill try to develop the core of an open systemwith well defined interfaces that will certainlyfurther evolve over the years. We thereforeview our Volcano Fast Response System as aseed for a Volcanic Task Force Team.We deem the inclusion of older geophysicaldata that have been recorded by the alreadyexisting network into the fast response systemdatabase extremely important because theyserve as a baseline for assessing the currentactivity. In case the Volcano Fast Response Sys-tem is deployed at a volcanic system wherethe past activity is fairly unknown, an essentialpart of the Task Force that operates the Vol-cano Fast Response System would involve athorough mapping of the past activity. Pastactivity is an essential information and one ofthe keys when trying to assess future activity(e.g. Newhall and Punongbayan, 1996). How-ever, this is beyond the scope of this project asthis depends on the individual volcano to bemonitored.

2.1. Project strategyThe project includes a total of 9 groups work-ing on different parts of the Volcanic FastResponse System. We therefore divided theproject into 5 work packages which containrelated tasks. This enhances communicationbetween the different groups working onrelated topics. Three of the work packagesfocus on advancing ground based (WP1) andspace based (WP2) observational techniquesas well as developing a central database, analert level system and visualization tools(WP3). One is focussing on quantitative, phys-ical models and data interpretation (WP5).

127

Finally WP4 hosts the development of thecommunication network and the prototypeinstallation of the VFRS as well as the overallcoordination of the whole project.

Work Package 1: Ground based observationsPIs: C. Gerstenecker, M. Becker (Darmstadt),T.H. Hansteen (Kiel)Goal of this work package is to provide novelinstrumentation for ground based observationsto the VFRS. The main problem of instrument-ing a volcano during a crisis isa) the availability of a significant number of

instruments that

Figure 1: Connection between the five different workpackages

b) can be installed at a safe distance to the vol-cano still providing key data which

c) all need to be transmitted in real time to acentral system

d) in order to be evaluated in real time.The most known and most established obser-vational method is the observation of seismicactivity. Broadband seismometers are welldeveloped and for our VFRS test installation(see WP4 below) we will use three componentbroadband instruments from the amphibianpool (DEPAS). Aside from seismic observationswe incorporate two novel ground basedobservational techniques: a) high resolutionlocal deformation measurements performedby a ground based InSAR system, and b)ground based gas measurements. The greatadvantage of ground based InSAR systems,which become commercially available now, istheir high resolution as well as repetitionmeasurements that are not tied to cycle timesof satellites, and the immediate availability ofthe data. Degassing rates are also a key meas-urement indicating activity changes in themagmatic system (see Fig. 2). We will there-fore incorporate a mini DOAS Instrument intoour system for continuous measurements ofgas concentrations and fluxes of various vol-canic volatiles.

Figure 2: Temporal changes of the SO2 emissions at San Cristobal volcano, Nicaragua, Nov. 23rd, 2002. Clearly visibleis the high temporal variability of the SO2 flux underpinning the importance of continuous gas flux measurement in theframework of the VFRS.

128

Work Package 2: Space based observationsPIs: N. Adam M. Eineder, T. Erbertseder,P. Valks, (DLR, Oberpfaffenhofen), W. Thomas(Offenbach), R. Bamler, S. Hinz (TU Mün-chen), M. Hort, D. Stammer (Hamburg)

Goal of this work package is to include spacebased observations into the VFRS because inrecent years developments in space basedobservations have more and more opened upa new field in volcanology through allowingthe areal mapping of deformation (Massonetteet al, 1993; satellites/instruments: Envisat,ERS1,2, ALOS, TerraSAR), the detection ofthermal spots (Wright et al, 2004,satellites/Instruments: AVHRR, GOES, MODIS,ASTER), and the identification of broaddegassing signatures, especially of SO2 (e.g.Eisinger and Burrows, 1998, satellites/instru-ments: GOME-2). Near real time data arealready becoming available with satellites hav-ing access times of 2 days (TerraSAR), plannedmissions getting down to 12 hours (FIRES), ordata from the geostationary system GOESbeing available every 15 min.The strategy for including space based obser-vations is twofold: a) we will use older satel-lite images to assess the evolution of defor-

mation and degassing before the crisis, and b)a real novelty of this mobile system is that wewill include the near realtime data from Ter-raSAR, GOME, MODIS which come in aboutevery two days. New data from GOME-2 datawill be available every day.

Work Package 3: Databases, IT-Architectureand VisualizationPIs: K. Klinge, K. Stammler (SZGRF – BGR),J. Wassermann (LMU-München)

Goal of this work package is to providethe data base including GIS capability andvisualization tools and the determination ofalert levels. The success of the fast responsesystem including early warnings lives throughits connection capability to already installedmonitoring systems and the new wirelessnetwork as well as through its combinationof different raw and model data. At presentnone of the existing databases are able tohandle the multi-parameter data resultingfrom modern volcano monitoring networkssimultaneously which are quite common today(e.g. Zschau et al. 1998; Neuberg, 2000;Richter et al., 2004) to assess volcanic activity.In practice this means a high-dimensional,

Figure 3: IT Architecture and exchange protocols of the »expert system«

129

complicated (raw or already parameterised)data stream with different sampling rates andtime histories that have to be stored andanalysed. The success of the system will beclosely tied to the systems capability to visual-ize the results of the data analysis as well asresults from model calculations. Part of thedata recorded (e.g. gas measurements) can bedisplayed directly whereas, for example, seis-mic data need to be processed to determinee.g. hypocenters (see WP5).All direct data as well as analysis results arestored in a GIS based meta data base andtools will be developed to either look at thedata in real time or to visualize the temporalevolution over a certain time period (seeFig. 3). Taking all recorded data as well as thetemporal evolution of the system into accountautomatic alert levels will be determined.By overlaying existing GIS systems on land use,vulnerable objects, and critical industrial com-plexes etc, existing evacuation plans as well asevacuation routes can be reconsidered in thelight of the current activity and the results ofmodel predictions.

Work Package 4: Prototype installation of theVolcano Fast Response System and overall pro-ject coordinationPIs: T. Dahm, M. Hort (Hamburg)

Goal of this work package is to develop thewireless network, install and test the VolcanoFast Response System (land and sea parts) andcoordinate the whole project. A key to the suc-cess of the volcano fast response system is itswireless network that allows the fast deploy-ment of additional instrumentation in additionto an existing network. The network will bebased on so called meshnodes which build aselforganized multipoint to multipoint net-work, which reroutes data via a new connec-tion once a single connection fails (see Fig. 4).The VFRS itself will be tested in a prototypeinstallation on the Azores because there is cur-rently an unclear situation at Fogo volcano onSao Miguel near Ponta Delgada. Fogo volcanois showing seismic unrest since May 2002 andGPS measurements indicate a deflation on theNE flanks of Fogo volcano. The test installationincludes a land and marine part that will be

Figure 4: General WLAN network design of the VFRS. Please note that the design is not specifically optimized for useduring the test experiment but is a general layout allowing installation in very different environments. The arrows indi-cate all different ways of routing the data in this multipoint network.

130

deployed at the same time. The installation ofthe VFRS (land part) is scheduled for early2009 because by that time the basic function-ality of the system will be available. The asso-ciated marine expedition will take place inMay 2009. Carrying out the installation muchlater would not leave enough time to improvethe system. This will be achieved throughreplaying the data collected during the proto-type installation in real time to further devel-op the functionality of the system.

Work Package 5: Multiparameter analysis ofcontinuous network data for quantitativephysical model buildingPIs: T. Dahm (Uni Hamburg), M. Ohrnberger(Uni Potsdam), Th. Walter (GFZ Potsdam),J. Wassermann (LMU Uni München),U. Wegler (Uni Leipzig)

Goal of this work package is to provide physi-cal models to the VFRS that allow the determi-nation of the state of stress inside the volcano.Understanding volcanic systems is fairly diffi-cult not only because they are build from aseries of intrusive and extrusive/explosiveevents but also because there are a lot of stillpoorly understood and/or unknown internalfeedback mechanisms. Structure as well asvarious internal feedback mechanisms mayfinally lead to a critical stress change in thesystem and culminating in an eruption. Thiswork package strongly builds on the seismo-

logical observations and deformation meas-urements in order to constrain the activitystatus of a volcano (see Fig. 5). In particularalgorithms for event detection and classifica-tion and inversion tools for transient andquasi-continuous signals will be developed.Locating the recorded events will be based ondifferent approaches including seismic mo-ment tensor inversion. These results will betransferred to WP3 in order to be displayed inreal time.

3. Deployment strategies for the VFRSA prerequisite for the temporary installation ofthe VFRS is the request of a foreign adminis-tration to our government for help and assis-tance to manage a period of volcanic unrestor a volcanic crisis. The request for help caninclude various levels of technical assistanceranging from simply lending instruments allthe way to the installation of different com-ponents of the VFRS. Before actually deployingparts or the whole systems it must be deter-mined in detail, what type of help is asked forby the foreign government and what can beprovided by us. This includes a detailed expla-nation by the coordinator of the VFRS whatsystems are available. At this point it will be ofutmost importance to adhere to the local gov-ernmental structures when discussing whattype of help is needed. This may require assist-ance from the local German embassy in orderto avoid any misunderstanding.

Figure 5: Scheme for near-real-time dislocation source and stress field modeling. Sketch demonstrates InSAR observa-tion of a deforming volcano (1), the deformation data of which (2) are then inverted in elastic dislocation models (3).The dislocation source obtained from the inversion shall then automatically be included in forward models (4) to calcu-late the static and static stress field change at regions of interest (magma chamber, faults or other heterogeneities).Illustrated here for data of Sierra Negra volcano, the Galapagos Islands (InSAR data courtesy of T. Walter (Potsdam) andF. Amelung (Miami)).

131

Once it is agreed upon which type of help canbe provided the coordinator puts together asmall team of scientists that will travel to pro-vide the help. The decision on who is going totravel will be based on a) specific requests ofthe foreign authorities who may have workedalready with geoscientists from Germany, b)knowledge of the instrumentation requested,and c) on the expertise depending on the typeof volcanic system being at unrest. The groupof people will be kept as small as possible inorder to not give the impression of overtakingthe local agencies. Scientists of the VFRS pro-vide all information gathered by the VFRS tothe local scientists. They will also teach localscientists on how to use the systems, so thatthey can keep using the technology once thecrisis is over if that is agreed upon betweenboth parties. During the whole mission of theVFRS participating scientists will not communi-cate any results and predictions of what maybe happening to the local as well as the inter-national agencies and the press in order to a)give the local authorities a change to speakwith one voice and b) avoid any confusion onwho is in charge in managing the crisis. Incharge of managing the crisis will always bethe local scientists and agencies. As many vol-canic crisis have shown before speaking withone voice is one of the keys to a successful cri-sis management.

4. Relation to the current socialdiscussionReducing the impact of natural disasters onthe human habitat is a fundamental researchgoal. In volcanology the ultimate objective isto predict volcanic eruptions in time andspace. Chances to succeed are high becausevolcanoes show many different types of pre-cursors that can often be recognized prior toan eruption. In addition volcanoes are knowpoint sources with a very well documentedhistory. Unfortunately most volcanoes, espe-cially in third world countries, are still unmon-itored. It is therefore highly desirable to pro-vide help in case of a volcanic crisis in order toreduce risk and fatalities. With the technologyand system developed in the framework of

this project Germany will be in a position tooffer help in case of a volcanic crisis, providedthat additional funding for instruments tooperate a task force is available. Offering suchhelp in case of a request is therefore an inter-nationally visible effort plus it would alsoinclude transfer of technology and knowledgeinto poorer countries. In addition to the taskforce of the USGS this would be the secondtask force worldwide and due to partially dif-ferent instrumentation both task forces wouldcomplement each other nicely.

ReferencesBarrientos SE, (1994) Large thrust earthquakesand volcanic eruptions. Pure Appl Geophys142: 225–237

Bonaccorso A, Calvari S, Garfi G, Lodato L,Patane D, (2003) Dynamics of the December2002 flank failure and tsunami at StromboliVolcano inferred by volcanological and geo-physical observations. Geophys Res Lett: 30,10.1029/2003GL017702

Borgia A, Delaney P, Denlinger RP, (2000)Spreading volcanoes. Ann Rev Earth Planet Sci28: 539–570

Crisp JA, (1984) Rates of magmatism. J Vol-canol Geotherm Res 20: 177–211

Eaton JP, Murata DJ, (1960) How volcanoesgrow. Science 132: 925–938

Eichelberger JC, (1995) Silicic volcanism;ascent of viscous magmas from crustal reser-voirs. Ann Rev Earth Planet Sci 23: 41–63

Eisinger M, Burrows JP, (1998) TroposphericSulfur Dioxide observed by the ERS-2 GOMEInstrument, Geophys Res Lett 25: 4177–4180

Emter D, (1997) Tidal triggering of earth-quakes and volcanic events. In: Lecture Notesin Earth Sciences 66: Tidal Phenomena, Wil-helm H, Zürn W, Wenzel HG (eds). SpringerBerlin Heidelberg New York, pp. 293–310

132

Ewert, JW, Miller CD, (1995), http://vul-can.wr.usgs.gov/Vdap/Publications/OFR95–553/OFR95–553.html

Hill DP, Pollitz F, Newhall C (2002) Earth-quake-volcano interaction. Physics Today Nov2002: 41–47

Holloway JR (1976) Fluids in the evolution ofgranitic magmas: consequences of finite CO2

solubility. Geol Soc Am Bull 87: 1513–1518

Hort M, Seyfried R, Vöge M (2003) RadarDoppler velocimetry of volcanic eruptions:theoretical considerations and quantitativedocumentation of changes in eruptive behav-iour at Stromboli volcano, Italy. Geophys J Int154: 515–532

Lipman PW, Mullineaux DR (eds) (1981) The1980 eruption of Mount St. Helens. US GeolSurv Prof Pap 1250: 1–844

Marzocchi W, Casarotti E, Piersanti A (2002)Modelling the stress variations induced bygreat earthquakes on the largest volcaniceruptions of the 20th century. J Geophys Res107, 2320, DOI: 10.1029/2001JB001391

Massonnet D, Rossi M, Carmona C, AdragnaF, Peltzer G, Feigl K, Rabaute T (2993) Thedisplacement field of the Landers earth-quake mapped by radar interferometry.Nature 364: 138–142

McCormick PM, Thomason LW, Trepte CR(1995) Atmospheric effects of the Mt.Pinatubo eruption. Nature 373: 399–404

Moran SC, Stihler SD, Power JA (2002) A tec-tonic earthquake sequence preceding theApril–May 1999 eruption of Shishaldin Vol-cano, Alaska. Bull Volcanol 64: 520–524

Neuberg J (2000) External modulation of vol-canic activity. Geophys J Int 142: 232–240

Newhall C, Punongbayan RS (1996) The nar-row margin of sucessful volcanic-risk mitiga-

tion. In Scarpa R, Tilling RI (eds) Monitoringand mitigation of volcanic hazards,Springer Verlag, pp. 807–838

Nostro C, Stein RS, Cocco M, Belardinelli ME,Marzocchi W (1998) Two-way couplingbetween Vesuvius eruptions and southernApennine earthquakes, Italy, by elastic stresstransfer. J Geophys Res 103: 24487–24424

Pallister JS, Hoblitt RP, Reyes AG (1992): Abasalt trigger for the 1991 eruptions ofPinatubo volcano. Nature 356: 426–428

Richter G, Wassermann J, Zimmer M, Ohrn-berger M (2004) Correlation of seismic activi-ty and fumarole temperature at Mt. Merapivolcano (Indonesia) in 2000. J Volcanol Geo-therm Res 135: 331–342

Schmincke HU (2004) Volcanism.Springer Verlag, Berlin.

Sigvaldason GE (1989) International confer-ence on Lake Nyos disaster, Yaoundé,Cameroon 16–20 March, 1987: Conclusionsand recommendations, J. Volcanol GeothermRes 39: 97–107

Simkin T, Siebert L (1994). Volcanoes of theWorld, 2nd edition. Geoscience Press, Tuc-son, 349 p.

Tilling RI, Dvorak JJ (1993) Anatomy of abasaltic volcano. Nature 363: 125–133

Tolstoy M, Vernon FL, Orcutt JA, Wyatt FK(2002) Breathing of the seafloor: tidal correla-tions of seismicity at Axial volcano. Geology30: 503–506

Turner JS, Campbell IH (1986) Convectionand mixing in magma chambers. Earth SciRev 23: 255–352

Violette S, de MG, Carbonnel JP, Goblet P,Ledoux E, Tijani SM, Vouille G (2001)Can rainfall trigger volcanic eruptions? Amechanical stress model of an active volcano;

133

»Piton de la Fournaise«, Reunion Island. TerraNova 13: 18–24

Walter TR, Acocella V, Neri M, Amelung F(2005) Feedback processes between volcanicactivity and flank slip at Mt. Etna (Italy)during the 2002–2003 eruption. Journalof Geophysical Research 110, B10205,doi:10.1029/2005JB003688

White JDL, Houghton B (2000) Surtseyan andrelated phreatomagmatic eruptions. In: Si-gurdsson H, Houghton BF, McNutt S, Rymer H,Stix J (eds) Encyclopedia of Volcanology. Aca-demic Press, San Diego, pp 495–511

Wright R, Flynn LP, Garbeil H, Harris AJL, PilgerE. (2004). MODVOLC: near-real-time thermalmonitoring of global volcanism. J VolcanolGeotherm Res 135: 29–49

Zschau, J. Sukhyar, R., M.A. Purbawinata,B.-G. Lühr; M. Westerhaus (1998): ProjectMERAPI – Interdisciplinary Research at a High-Risk Volcano; DGG-Mitteilungen, SonderbandIII/1998, S. 3–8

134

Author’s Index

AAbecker A. . . . . . . . . . . . . . . . . . . 113Arnhardt C. . . . . . . . . . . . . . . . . . . 75Asch K. . . . . . . . . . . . . . . . . . . . . . . 75Azzam R. . . . . . . . . . . . . . . . . . . . . 75

BBecker R. . . . . . . . . . . . . . . . . . . . . 89Bell R. . . . . . . . . . . . . . . . . . . . . . . . 89Bill, R. . . . . . . . . . . . . . . . . . . . . . . . 75Birkmann . . . . . . . . . . . . . . . . . . . . 62Boley C. . . . . . . . . . . . . . . . . . . . . 113Bonn G. . . . . . . . . . . . . . . . . . . . . . 31Breunig M. . . . . . . . . . . . . . . . . . . 113Buchmann A. . . . . . . . . . . . . . . . . . 31Burghaus S. . . . . . . . . . . . . . . . . . . 89

DDahm T. . . . . . . . . . . . . . . . . . 14, 124Danscheid M. . . . . . . . . . . . . . . . . . 89Dech . . . . . . . . . . . . . . . . . . . . . . . . 62Dix A. . . . . . . . . . . . . . . . . . . . . . . . 89

EErdik M. . . . . . . . . . . . . . . . . . . . . . 51

FFernandez-Steeger T. M. . . . . . . . . . 75Fischer J. . . . . . . . . . . . . . . . . . . . . . 41Friederich W. . . . . . . . . . . . . . . . . . . 14

GGalas R. . . . . . . . . . . . . . . . . . . . . . . 7Gallus D. . . . . . . . . . . . . . . . . . . . . 113Ge M. . . . . . . . . . . . . . . . . . . . . . . . . 7Gendt G. . . . . . . . . . . . . . . . . . . . . . 7

Glade T. . . . . . . . . . . . . . . . . . . . . . 89Greiving S. . . . . . . . . . . . . . . . . . . . 89Greve K. . . . . . . . . . . . . . . . . . . . . . 89Gurgel K.-W. . . . . . . . . . . . . . . . . . . 20

HHanka W. . . . . . . . . . . . . . . . . . . . . 14Helzel T. . . . . . . . . . . . . . . . . . . . . . 20Heunecke O. . . . . . . . . . . . . . . . . . 101Hilbring D. . . . . . . . . . . . . . . . . . . . 31Hirzinger . . . . . . . . . . . . . . . . . . . . . 62Hohnecker E. . . . . . . . . . . . . . . . . . 31Homfeld S. D. . . . . . . . . . . . . . . . . . 75Hort M. . . . . . . . . . . . . . . . . . . . . . 124

JJäger S. . . . . . . . . . . . . . . . . . . . . . . 89

KKallash A. . . . . . . . . . . . . . . . . . . . . 75Kazakos W. . . . . . . . . . . . . . . . . . . 113Kind R. . . . . . . . . . . . . . . . . . . . . . . 14Klein . . . . . . . . . . . . . . . . . . . . . . . . 62Klüpfel . . . . . . . . . . . . . . . . . . . . . . 62Kniephoff M. . . . . . . . . . . . . . . . . . 20Krüger F. . . . . . . . . . . . . . . . . . . . . . 14Krummel H. . . . . . . . . . . . . . . . . . . 89Kühler T. . . . . . . . . . . . . . . . . . . . . . 31Kuhlmann H. . . . . . . . . . . . . . . . . . 89

LLehmann . . . . . . . . . . . . . . . . . . . . 62Lessing R. . . . . . . . . . . . . . . . . . . . . 51Lupp M. . . . . . . . . . . . . . . . . . . . . . 51

MMäs S. . . . . . . . . . . . . . . . . . . . . . 113Meier T. . . . . . . . . . . . . . . . . . . . . . 14Milkereit C. . . . . . . . . . . . . . . . . . . . 51Mott . . . . . . . . . . . . . . . . . . . . . . . . 62

NNagel . . . . . . . . . . . . . . . . . . . . . . . 62Niemeyer F. . . . . . . . . . . . . . . . . . . . 75

OOhrnberger M. . . . . . . . . . . . . . . . . 14Ortlieb E. . . . . . . . . . . . . . . . . . . . 113

PPaulsen H. . . . . . . . . . . . . . . . . . . . . 89Pohl J. . . . . . . . . . . . . . . . . . . . . . . . 89Pohlmann T. . . . . . . . . . . . . . . . . . . 20

QQuante F. . . . . . . . . . . . . . . . . . . . . 31

RRedlich J. P. . . . . . . . . . . . . . . . . . . . 51Reinhardt W. . . . . . . . . . . . . . . . . . 113Richter D. . . . . . . . . . . . . . . . . . . . 113Ritter H. . . . . . . . . . . . . . . . . . . . . . 75Röhrs M. . . . . . . . . . . . . . . . . . . . . . 89Rothacher M. . . . . . . . . . . . . . . . . . . 7

SSchedel F. . . . . . . . . . . . . . . . . . . . . 31Scherbaum F. . . . . . . . . . . . . . . . . . 14Schlick T. . . . . . . . . . . . . . . . . . . . . . 20Schlurmann . . . . . . . . . . . . . . . . . . 62Schöbinger F. . . . . . . . . . . . . . . . . . 31Schöne T. . . . . . . . . . . . . . . . . . . . . . 7Schubert C. . . . . . . . . . . . . . . . . . . . 51Setiadi . . . . . . . . . . . . . . . . . . . . . . 62Siegert . . . . . . . . . . . . . . . . . . . . . . 62Stammer D. . . . . . . . . . . . . . . . . . . 20Stammler K. . . . . . . . . . . . . . . . . . . 14Strunz . . . . . . . . . . . . . . . . . . . . . . . 62

TThuro K. . . . . . . . . . . . . . . . . . . . . 101Toloczyki M. . . . . . . . . . . . . . . . . . . 75Trauner F. X. . . . . . . . . . . . . . . . . . 113

WWalter K. . . . . . . . . . . . . . . . . . . . . 75Wassermann J. . . . . . . . . . . . . . . . 124Wenzel F. . . . . . . . . . . . . . . . . . 31, 51Wiesel J. . . . . . . . . . . . . . . . . . . . . 113Wunderlich T. . . . . . . . . . . . . . . . . 101

YYuan X. . . . . . . . . . . . . . . . . . . . . . . 14

ZZschau J. . . . . . . . . . . . . . . . . . . . . . 51

135

136

GEOTECHNOLOGIEN Science Report’s –Already published/Editions

No. 1 Gas Hydrates in the Geosystem – Sta-tus Seminar, GEOMAR Research CentreKiel, 6–7 May 2002, Programme &Abstracts, 151 pages.

No. 2 Information Systems in Earth Mana-gement – Kick-Off-Meeting, Universityof Hannover, 19 February 2003,Projects, 65 pages.

No. 3 Observation of the System Earth fromSpace – Status Seminar, BLVA Munich,12–13 June 2003, Programme &Abstracts, 199 pages.

No. 4 Information Systems in Earth Mana-gement – Status Seminar, RWTHAachen University, 23–24 March 2004,Programme & Abstracts, 100 pages.

No. 5 Continental Margins – Earth’s FocalPoints of Usage and Hazard Potential –Status Seminar, GeoForschungsZen-trum (GFZ) Potsdam, 9–10 June 2005,Programme & Abstracts, 112 pages.

No. 6 Investigation, Utilization and Protec-tion of the Underground – CO2-Stor-age in Geological Formations, Tech-nologies for an Underground SurveyAreas – Kick-Off-Meeting, Bundesan-stalt für Geowissenschaften undRohstoffe (BGR) Hannover, 22–23 Sep-tember 2005, Programme & Abstracts,144 pages.

No. 7 Gas Hydrates in the Geosystem – TheGerman National Research Program-me on Gas Hydrates, Results from theFirst Funding Period (2001–2004), 219pages.

No. 8 Information Systems in Earth Mana-gement – From Science to Application,Results from the First Funding Period(2002–2005), 103 pages.

No. 9 1. French-German Symposium on Geo-logical Storage of CO2, Juni 21./22.2007, GeoForschungsZentrum Pots-dam, Abstracts, 202 pages.





No. 10


ISSN: 1619-7399

In addition to currently implemented measures for establishing an early tsunamiwarning system in the Indian Ocean, the German Federal Ministry of Education andResearch (BMBF) has launched a portfolio of 11 research projects for developingand testing early warning systems for other natural geological catastrophes. Theprojects are carried out under the umbrella of the national R&D-Programme GEO-TECHNOLOGIEN.

The overall aim of the integrated projects is the development and deployment ofintegral systems in which terrestrial observation and measurement networks arecoupled with satellite remote sensing techniques and interoperable informationsystems. All projects are carried out in strong collaboration between universities,research institutes and small/medium sized enterprises on a national and interna-tional level.

The abstract volume contains the presentations given at the “Kick-Off-Meeting”held in Karlsruhe, Germany, in October, 2007. The presentations reflect the multi-disciplinary approach of the programme and offer a comprehensive insight into thewide range of research opportunities and applications.

The GEOTECHNOLOGIEN programme is funded by the Federal Ministry for

Education and Research (BMBF) and the German Research Council (DFG)

No.

10Ea

rlyW

arni

ngSy

stem

sin

Earth

Man

agem

ent

GEO

TECH

NO

LOG

IEN

Scie

nce

Repo

rt

Umschlag_SR10.qxd 20.09.2007 15:24 Uhr Seite 1

SR10

Documents

Transcript of SR10