Near Real-Time Geo-Analyses for Emergency Support:
A Radiation Safety Exercise
Guenther Sagl, Michael Lippautz, Bernd Resch, Manfred Mittlboeck
Studio iSPACE, Research Studios Austria
INTRODUCTION
A variety of well-established emergency support systems are employed for measuring and
monitoring potential physical and chemical hazards such as radioactive radiation or toxic gases.
Within such systems real-time data from in-situ geo-sensor networks play a crucial role. The
exchange of real-time data and, indeed, their integration into heterogeneous systems across different
rescue organizations such as fire-fighters or military are significant time critical factors in the decision
making process. Proprietary data formats and interfaces are often the major problem within this
integration. We therefore use open and international standards in an entirely service oriented
workflow, from sensor-data acquisition to visualisation and dissemination of newly generated
information.
Previous work in the field of geo-processing based on the Open Geospatial Consortium (OGC) Web
Processing Service (WPS) includes, for example, application related to water quality (Ninsawat et al.,
2008), bomb threat scenarios (Stollberg and Zipf, 2007) etc. Although the current WPS 1.0.0
specification (Schut, 2007) has clear weak-points, for example its unlimited scope and lack on
restrictions (Michaelis and Ames, 2009; Stollberg and Zipf, 2007),or missing asynchronous
processing (Resch et al., 2010b), it is the only established web-based processing standard in the geospatial domain. We therefore implement only mandatory elements of the OGC WPS 1.0.0. Interface.
In this paper we demonstrate a real-time or near real-time – summarized by the term ‘live’ –
workflow for geo-sensor information analysis based on open standards in order to enhance timecritical emergency support. Emphasis is put on live geo-processing and rapid information
dissemination and visualisation. This workflow has been successfully applied as part of the ‘G2real’
project exercise ‘Shining Garden’ in Seibersdorf, Austria.
Geography to Reality – G2real
The overall aim of the FP6 ERA-STAR Regions project “G2real: Galileo based GMES1 real time
emergency support testbed, real time exercise and development of services” is the development and
test of new pre-operational GMES services in the field of emergency and disaster management. The
primary project objective is the validation and verification of the services developed through two realtime exercises – one relates to Galileo Navigation, the other to GMES services. Herein, we show
scientific results of the latter exercise for radiation safety called ‘Shining Garden’.
The G2real ‘Shining Garden’ exercise has been accomplished within a safe environment at
Seibersdorf Laboratories in Seibersdorf, Austria. Two 137Cs radiation sources were placed for
localisation. Subsequently, a person was moving through the test area sensing equivalent radiation
dose rate with an intelligent sensor pod (Figure 2). At the same time, spatial interpolation results of
these live sensor measurements were growing gradually and were visualized simultaneously on
several clients.
1
Global Monitoring for Environment and Safety
METHODOLOGY
The overall workflow design as illustrated by Figure 1 is based on the ‘Live Geography’ approach
introduced by Resch et al. (2009). We apply this workflow in order to maximize interoperability in
using well-established OGC standards. Specifically, we employ Sensor Observation Service (SOS) to
request live sensor measurements (Figure 1 ‘Data’), WPS (Schut, 2007) to transform data into
information (Figure 1 ‘Processing and Simulation’), and Web Feature Service (WFS), Web Coverage
Service (WCS), and Web Map Service (WMS), respectively, to rapidly disseminate and visualize
analysis results for browser- and web-based platforms (Figure 1 ‘Presentation’).
Figure 1: modular and service-oriented workflow from data (left) to
presentation of information (right) based on standardized interfaces
Data: From Measurements to Sensor Observation Service
A variety of high-quality sensing devices have the potential to serve as real-time data sources.
Quantitative properties of the physical or chemical phenomenon – in this case dose rate of radioactive
radiation – need to be measured by an accurately calibrated sensors. The measurements needs to be
pre-filtered, tagged with its current spatial position and time-stamp, and finally published as SOS for
further use within the service-oriented infrastructure (Figure 1). The sensor pod framework developed
is fully compatible with already existing sensors. As shown in Figure 2, we use a separated embedded
system, the IGEPv22 single board computer, as central managing unit which manages data received
from a sensing device (e.g. SSM-1), and a Global Positioning System (GPS) receiver. Both sensors
transmit raw data via serial interfaces, e.g. RS232 (SSM-1) or USB (GPS), to a measuring framework
running on the embedded computer.
2
www.igep.es
In essence, the measuring framework on the embedded computer manages quality assurance
routines of raw measurements. These routines include filtering out non-fixed GPS positions as well as
malformed values. After this first verification step the data are stored in a SQLite3 embedded
database. The database is used to decouple the measurement phase from further processing steps.
Additionally, the sensing framework and the database are set up to act in a round-robin scheduling
manner. This very loose coupling enables concurrent actions such as measuring, and delayed data
delivery using a HTTP service. The SOS interface on the integrated webserver is then used for further
geo-processing and analyses steps (see subsection Geo-Processing). The services provided currently
include SOS, Keyhole Markup Language (KML), and Geo Really Simple Syndication (GeoRSS). The
entire functionality of this sensor pod approach is conform with OGC Sensor Web Enablement
(SWE) (Botts et al., 2007) and is described in detail by Resch et al. (2010a).
Figure 2: Sensor Pod: GPS receiver (1), IGEPV2 Single Board Computer (2),
and UMTS modem (3) on top of SSM-1 (circuit board only)
Data Integration
Since the recent emergence of a variety of real-time data sources the ‘topicality’ parameter, i.e. the
up-to-dateness, in addition to other quality criteria (accuracy, completeness etc.), is receiving a lot of
attention. Although such real-time geo-data enhance spatial analyses, these are indispensible for timecritical decision support. The integration of real-time data into decision support systems, however,
often requires pre-processing steps such as format conversions. Thus, a data fusion mechanism for
live integration of real-time data is required. We therefore developed ‘live-data-source-plug-ins’ for
open source and commercial software packages.
3
www.sqlite.org
The ‘live-data-source-plug-in’ enables the direct integration of SOS conform data structures into
GIS applications. The open source software GeoServer4 recognises this plug-in as a ‘data-store’ and
converts the SOS input ‘on-the-fly’ into a variety of output formats including KML, GeoRSS, PDF,
and SVG (beside WFS, WMS, WCS). ESRI ArcGIS as a commercial GIS product recognises this
plug-in as a ‘simple point feature class’. This point layer contains live measurements which can be
used for further spatial as well as temporal analysis. As a result, real-time data fusion as described
above enables the direct integration of live measurements into a variety of systems, not necessarily GI
Systems.
Geo-Processing
Discrete live in-situ measurements are spatially represented as points. To convert these onedimensional – geometrically zero-dimensional – measurements of continuous phenomena (e.g. dose
rate of radioactive radiation) into multi-dimensional information layers we employ IDW as a
deterministic-, and Kriging as a stochastic spatial interpolation technique. We therefore developed an
integrated modular geo-processing workflow. The workflow’s three basic components are illustrated
by Figure 3: (1) transformations of input data from its inherent spatial reference system (e.g. WGS84)
to a projected coordinate system (e.g. UTM 33) – further metric calculations require a lengthpreserving map; spatial interpolation of projected point data. Additional calculations, if necessary, are
assigned to this component; (2) spatial interpolation of points to a continuous surface. The technique
of interpolation (IDW, or Kriging) as well as its parameter (e.g. exponent of distance for IDW) is
selectable and adjustable at run-time5 by the user; (3) classification (e.g. low, moderate, and high) of
processing results according to user-specific thresholds. The first two components relate to ‘pure’
geo-processing. The third component reduces information to its essence for usability in time-critical
decisions.
Figure 3: principal modular live geo-processing workflow
Based on that workflow, and in compliance with SOA, we establish geo-processing services as
described by Mittlboeck et al. (2010). ESRI ArcGIS Server6 in combination with PyWPS7 acts as the
live geo-processing engine. In the near future, due to the massive increase of real-time data,
comprehensive architectures for distributed and cloud processing might be required (Friis-Christensen
et al., 2007; Schaeffer et al., 2009). Depending on the format requested by the user, the geoprocessing output includes vector or raster data. This results in an entirely service oriented live geoprocessing workflow utilizing the following OGC standards: SOS to request one-dimensional in-situ
data, WPS for geo-processing on-the-fly, and WFS, WCS, and WMS, respectively, for multidimensional information output (Figure 3 bottom).
4
www.geoserver.org
changes take effect when a new geo-process task is triggered (running tasks are not affected)
6
www.esri.com
7
http://pywps.wald.intevation.org
5
Presentation: Information Dissemination and Visualization
Correct and easily interpretable visualisation of complex information is a central aspect in timecritical decision making processes. We prepare such information as standardized services and enable
their rapid dissemination within internet-based environments. These services can therefore be easily
integrated into other applications including GI Systems for enhanced visualisation.
In order to improve the situational awareness of decision makers, and even rescue teams on-site, we
focus on widely accepted visualisation clients including Google Earth. Usability in terms of
recognition of other browser-based user interfaces is of significant importance, especially when
enhanced analysis capabilities are provided. We therefore emphasise on simple user interface design
principles for both general and application-specific websites.
RESULTS
Results shown herein are outcomes of the ‘Shining Garden’ exercise. As mentioned in the
introduction, two 137Cs radiation sources were placed for localisation. Prior to the exercise we
measured a radiation dose rate of 6.3 µ Sievert per hour (µSv/h) (western source), and 4.7 µSv/h
(eastern source).
Web-Based Visualisation
Figure 4 shows a screenshot of the geo-processing web-application developed for the ‘Shining
Garden’ exercise. The background within that screenshot shows a 2010 GeoEye satellite imagery
integrated as a WMS. The radiation safety test area is located in the centre of the imagery (Figure 4
red frame).
Figure 4: Screenshot of the G2real web-application (red frame: radiation saftey test area)
The graphical user interface requires only two user interactions to trigger live analysis: (1) numeric
entry of the update time-interval (in seconds) for the re-triggering of geo-processing; (2) selection of
the preferred interpolation method (IDW or Kriging) and its output visualisation format (isolines with
user defined step size, classified isoareas, or continuous surface). This selection can be changed by
the user between the idle intervals of processing.
Figure 5 illustrates IDW (exponent of distance: 2) interpolation results in relative chronological
order (1–6). In analogous manner, Figure 6 show results based on Ordinary Kriging interpolation
(semi-variogram model used: spherical). The violet points represent discrete locations of sensor
measurements and thus the path taken by the person carrying the sensor pod. A proposed and
significant difference between both figures is the path taken by the person carrying the sensing
device.
Figure 5: Screenshots of live IDW interpolation (green < 0.5 µSv/h; gradient from green to red: >=
0.5…<= 3µSv/h; red: > 3µSv/h); points represent discrete sensor measurement locations
Figure 6: Screenshots of live Kriging interpolation (green < 0.5 µSv/h; gradient from green to red:
>= 0.5…<= 3µSv/h; red: > 3µSv/h); points represent discrete sensor measurement locations
Post-Processing for In-Depth Analysis
Figure 7 shows interpolation results in a matrix manner: columns represent the interpolation method
applied (IDW, and Kriging); rows represent the first and second phase, respectively, of the ‘Shining
Garden’ exercise. These results has been generated after the exercise but are based on the same
measurements as results presented in Figure 5 and Figure 6.
Figure 7: Comparison of interpolation results: IDW versus Kriging in Phase 1 and Phase 2
DISCUSSION
The overall workflow presented has been successfully verified in the course of the “Shining
Garden” real-time exercise on radiation safety. Interpolation results shown in Figure 5 and Figure 6
demonstrate gradually increasing spatial awareness of radiation dose rate in a live manner. The
figures show the path taken by the person (e.g. used for reporting purposes) as combination of
discrete sensor measurements represented as points, and their spatial interpolation results growing
successively. Stepwise interpolation results shown in Figure 5 (1-6) indicate a correct and spatially
explicit detection of radiation sources during the sensing scenario. The final interpolation result
shown in screenshot 6, however, shows a considerable distortion in the south-western area. This is
because of un-sampled locations within that area in combination with the interpolation method used
(IDW). In contrast to Figure 5, intermediary interpolation results illustrated by Figure 6 indicate
higher uncertainty in terms of spatial variability, especially by comparing red areas in screenshot 2
and 3, and 4 and 5. The final result (screenshot 6 in Figure 6) nevertheless shows a spatially exact
localisation of both radiation sources displayed for detection. The direct comparison of final
interpolation results shown in Figure 7 takes the two different paths into account. It clearly indicates
for both phases that Kriging rather than IDW fulfils an accurate localisation of radiation sources.
The phenomenon measured – radiation dose rate – has a favourable characteristic with respect to
sensing: it has, within the time frame of the entire exercise (~ 3 hours), a constant spatial and
temporal variability – a fix positions of the radiation source is assumed. Thus, the latency within insitu measurements observed with ‘only’ one sensor device at different locations does not affect
succeeding spatial interpolation. This aspect, however, must be taken into account when sensing
phenomena with high spatiotemporal variability, for example the concentration of toxic gases.
CONCLUSION AND OUTLOOK
In this paper we demonstrated a verified live workflow for geo-sensor information analysis to
enhance time-critical emergency support. The workflow is based on open and international standards
for sensor-data acquisition (SOS), geo-processing (WPS), and visualisation and dissemination of
newly generated information (WMS, WFS, WCS). Geo-analyses results show that IDW seems to be
appropriate for interpolating measurements during the sensing process. After this process is finished
Kriging should be used to get the most suitable spatial assessment of radioactive radiation. This is in
agreement with the outcomes of the study performed by Mabit and Bernard (2007).
We conclude that the presented standards-based workflow significantly improves information
exchange for time-critical spatial decision support in terms of interoperability. In comparison to
previous research, we integrate up-to-date measurements from highly mobile intelligent sensor pods
on-the-fly into a fully service-oriented live geo-analysis workflow. Thus, with the given approach we
are able to process dynamic measurements in addition to mostly static legacy geodata.
In addition to the SOS used in this research, the Sensor Alert Service (SAS), and the Web
Notification Service (WNS) are OGC interfaces that are highly relevant for real-time emergency
support. In contrast to the pull-based OGC SOS, the SAS is based on the push principal, i.e. the
sensor itself sends information triggered by an event – a detailed discussion is given by Resch et al.
(2010a). This push-based service in combination with the asynchronous OGC WNS can enable
Complex Event Processing routines for the spatial domain, for example trigger further geo-processing
operations. In a new Sensor Web Enablement generation, eventing and alerting mechanisms will be
covered by the Sensor Event Service (Bröring et al., 2011).
Further research will incorporate above mentioned eventing and alerting mechanisms for real-time
decision support. Furthermore it will focus on performance issues of the geo-processing phase. In
addition to the allocation of distributed and cloud processing capabilities, the data’s intrinsic
spatiotemporal dynamics needs to be considered. So far, all available measurements at a given timestamp or period, and within a certain spatial extent, serve as input for analyses, independent if some
values remain constant since the last analysis. The increasing volume of sensor measurement
information, however, requires an effective and efficient mechanism for change detection of spatial
temporal parameter that exceeds variable thresholds. Consequently, selected partitions of a previous
analysis can be updated by fractional interpolation results considering significant changes only.
ACKNOWLEDGMENT
We thank all G2real project members for their contribution. Specifically, we thank Seibersdorf
Laboratories for the organisation of the ‘Shining Garden’ exercise. This research work is partially
founded by the Austrian Federal Ministry for Science and Research.
REFERENCES
Botts, M., Percivall, G., Reed, C. & Davidson, J. 2007. OGC® Sensor Web Enablement: Overview
And High Level Architecture Version 3. OGC White Paper.
Bröring, A., Echterhoff, J., Jirka, S., Simonis, I., Everding, T., Stasch, C., Liang, S. & Lemmens, R.
2011. New Generation Sensor Web Enablement. Sensors, 11 (3), 2652-2699.
Friis-Christensen, A., Ostländer, N., Lutz, M. & Bernard, L. 2007. Designing Service Architectures
for Distributed Geoprocessing: Challenges and Future Directions. Transactions in GIS, 11 (6),
799-818.
Mabit, L. & Bernard, C. 2007. Assessment of spatial distribution of fallout radionuclides through
geostatistics concept. Journal of Environmental Radioactivity, 97 (2-3), 206-219.
Michaelis, C. D. & Ames, D. P. 2009. Evaluation and Implementation of the OGC Web Processing
Service for Use in Client-Side GIS. Geoinformatica, 13, 109–120.
Mittlboeck, M., Resch, B., Lippautz, M. & Sagl, G. 2010. G2Real: ‚live‘ Geo(mess)information für
mobile Einsatzkräfte. In: STROBL, J., BLASCHKE, T. & GRIESEBNER, G. (eds.)
Angewandte Geoinformatik AGIT 2010. Salzburg, Austria: Wichmann Verlag, Heidelberg.
Ninsawat, S., Raghavan, V. & Masumoto, S. 2008. Integration of Web Processing Service and Sensor
Observation Service for distributed geoprocessing using real-time data. Geoinformatics, 13
(3).
Resch, B., Mittlboeck, M., Girardin, F., Britter, R. & Ratti, C. 2009. Live Geography – Embedded
Sensing for Standardised Urban Environmental Monitoring. International Journal on
Advances in Systems and Measurements, 2 (2&3), 156-167.
Resch, B., Mittlboeck, M. & Lippautz, M. 2010a. Pervasive Monitoring—An Intelligent Sensor Pod
Approach for Standardised Measurement Infrastructures. Sensors, 10, 11440-11467.
Resch, B., Sagl, G., Blaschke, T. & Mittlboeck, M. 2010b. Distributed Web-processing for
Ubiquitous Information Services – OGC WPS Critically Revisited. In: Proceedings of the 6th
International Conference on Geographic Information Science (GIScience2010), 14 - 17
September 2010, Zurich, Switzerland.
Schaeffer, B., Baranski, B., Foerster, T. & Brauner, J. 2009. A Service-Oriented Framework for Realtime and Distributed Geoprocessing. International Open Source Geospatial Research
Symposium OGRS 2009. Nantes, France.
Schut, P. 2007. OpenGIS® Web Processing Service. OpenGIS Standard, Version 1.0.0, OGC 05007r7 [Online]. Available: http://www.opengeospatial.org.
Stollberg, B. & Zipf, A. 2007. OGC Web Processing Service Interface for Web Service Orchestration
Aggregating Geo-processing Services in a Bomb Threat Scenario. In: WARE, J. & TAYLOR,
G. (eds.) Web and Wireless Geographical Information Systems. Springer Berlin / Heidelberg.