Slope Stability 2011: International Symposium on Rock Slope Stability in Open Pit Mining and Civil Engineering,

Vancouver, Canada (September 18-21, 2011)

Towards a Low Cost 3D Early Warning System for Unstable Alpine

Slopes The Aggenalm Landslide Monitoring System
K. Thuro Technische Universitt Mnchen, Chair for Engineering Geology, Munich, Germany
J. Singer Technische Universitt Mnchen, Chair for Engineering Geology, Munich, Germany
J. Festl Technische Universitt Mnchen, Chair for Engineering Geology, Munich, Germany

Abstract
A low cost landslide monitoring and early warning system has been developed as part of the alpEWAS project
over the last 3 years (2007-2010), where the goal was to develop and test a innovative, continuous monitoring
system. This system consists of time domain reflectometry (TDR) for the detection of subsurface displacements in
boreholes, reflectorless video tacheometry (VTPS) and a GNSS sensor network for the determination of 3D
surface movements. The latter have been integrated to provide real time data about trigger factors, such as
precipitation, which can be accessed remotely by a web interface. The Aggenalm Landslide, near Bayrischzell in
the Bavarian Alps was selected as the field laboratory for the project. The system has been operating in a
continuous mode for the last 2 years, with the exception of several minor disruptions, to acquisition data logging
components.

Introduction

The aim of the joint research project alpEWAS (Development and testing of an integrative 3D early warning
system for unstable Alpine slopes) was to derive innovative, efficient and cost-effective techniques for landslide
monitoring using a geo sensor network. The project was funded by the geoscientific research and development
program Geotechnologien of the German Federal Ministry of Education and Research (BMBF).
The need for a efficient and cost effective monitoring system is magnified by the apparent increase in landslides,
particularly in susceptible slopes that have not failed or their distribution is sporadic, making it difficult to
monitor with traditional approaches.
With the given option of a continuous remote access to the early warning system the stakeholder has the
possibility to view and control all incoming data from a central point, which allows simultaneous landslide
events, particularly during heavy rainfall to be evaluated simultaneously.
The project therefore focused on testing, integrating and advancing the following three measurement techniques
within an early warning system (Figure 1)

Time domain reflectometry (TDR)

Reflectorless video tacheometry (VTPS)

Real-time, low-cost GNSS sensors (LC GNSS)

Although details of the project and geosensor network have been published elsewhere, see for example Thuro et
al. (2009, 2010) and Singer et al. (2009c), the five main phases of the project are outlined here:
1. Selection and investigation of a project area.
2. Design and installation of the early warning system.
3. Learning phase of the early warning system.

4. Testing phase of the early warning system.

5. Automation and optimization of the early warning system.
Further information is also available at www.alpewas.de

Figure 1.

Schematic depiction of the alpEWAS geo sensor network installed at the Aggenalm Landslide.

In close collaboration with the Bavarian Environment Agency (Bayerisches Landesamt fr Umwelt, BLfU) the
Aggenalm Landslide was chosen as the test site since it consists of appropriate characteristics for the testing of
new technologies under alpine conditions. Since the present day movement rates of 1 cm/a were smaller than
expected at the initiation of the project, assessment of threshold values were affected or limited (project phase 2
and 3), for the testing of a monitoring system where a low hazard potential exists. Within the scope of the project
software applications for the sensor control, data management, complex data processing and integrative data
analysis were developed. After expiry of funding the monitoring of the Aggenalm Landslide will be continued
by the involved institutes.

Measurement techniques

2.1

Time Domain Reflectometry (TDR)

Quantitative measurements with TDR are influenced by several different parameters, which include the type of
cable used (lead and measurement cable) and the grout composition (e.g. cement, bentonite, cement additives).
These parameters were tested through a series of calibration tests (see Singer et al. 2009b, Singer et al. 2010,
Singer 2010).
The results of the laboratory tests combined with experience from field surveys of the TDR system were
incorporated into an installation handbook, which gives standardized installation recommendations consisting of
a combination of cable type and grout composition for different geological settings and types of landslide
phenomena (e.g. mechanisms and velocities). In addition to the installation handbook, deformation analysis
software for TDR was additionally developed.
The TDR system, that has been installed at the Aggenalm Landslide, has been operating continuously since
November 2008 with an hourly acquisition regime (installation and commencement of operation in September
2008). Over the total period since the beginning of operation (28 months), total data loss summed up to about
18 % of the planned measurements mostly because of power shortages and late data retrieval (full data storage
before the start of the automatic data acquisition). For the time since automatic data retrieval was possible data
loss added up to 10 %. Here, a major annoyance in January 2010 accounts for the loss of data. It is assumed that
in future greater reliability could be achieved due to the automatic data retrieval and data storage, as well as the
recently developed Status Monitor Software, as shown by results obtained since February 2010.
The field test at the Aggenalm has shown that most of the cables installed both surface and subsurface (lead and
measuring cables, connectors) proved to be robust concerning the weather, also during winter. To date, no
significant deformations have been measured with the TDR system at the Aggenalm Landslide, therefore the
signal and deformation analysis could only be used and tested during the evaluation of laboratory tests.
Figure 2 shows the results of the TDR measurements at the KB1 node, a mountain pasture location called
Lampl-Alm. Depicted are the reflection coefficient depth curves as a time series in relation to the reference
measurement in October 2008 as well as the results of the inclinometer measurements. The TDR time series
shows no measureable deformations, the reflection coefficient doesnt change at all during the depicted time
period. The results of the inclinometer measurements show slight movements (max. 1-2 mm for KB1). The
difference between the inclinometer results and the movements measured by the GNSS component is due to the
fact that technical problems arose while establishing KB1; with a depth of about 24,5 m the probable shear zone
wasnt reached. The TDR system, which was installed at a greater depth parallel to KB1, still doesnt show any
significant deformations. This may be explained because of the small amount of deformation has not yet been
able to fracture the grout surrounding the coaxial cable, what is a prerequisite for a deformation analysis with
TDR.

Figure 2.

2.2

Time series of the inclinometer measurements (left)

and the TDR measurements (above) at the sensor
node KB1. The inclinometer measurements
compared to the reference measurements show some
small insignificant variations of about 1-2 mm. Over
a time span of 70 weeks no deformations (variation
of reflection coefficient) were detected with TDR
measurements.

Reflectorless Video Tacheometry (VTPS)

One of the goals of the alpEWAS project was to gain practical field experience with an innovative instrument
not commonly used for displacement monitoring, known as video tacheometer (VTPS video tacheometric
positioning system). Currently the Topcon Imaging Station is the only instrument available on the market which
in principle has a video functionality with the necessary precision (Topcon Positioning Systems, Inc. 2008).
Other instruments the Trimble Spatial Station VX and the Leica Viva TS15 could be used only with major
limitations due to their low image resolution and restricted field of view (Trimble Navigation Limited 2010,
Leica Geosystems 2010). All of the aforementioned instruments cannot be used scientifically, as the camera
output cannot be read out by external devices. External camera control is only possible with a prototype system
from Leica Geosystems. It consists of a TPS1200 tacheometer with access to all instrument subsystems, which
has been modified with an internal camera (known as IATS2 Image Assisted Total Station 2nd edition).
Various design limitations, especially regarding practical handling, need to be addressed by the manufacturer.
Among others, this includes the tentative cable guiding and the taring of the modified telescope layout. The
prototype is described in Wasmeier (2009b) together with its extensive calibration procedure in Wasmeier
(2009a). Accuracy of the system was studied in the field revealing errors of approximately < 1 mgon on
average. Under poor conditions (high atmospheric refraction) up to > 2 mgon, whilst in the laboratory, an
accuracy less than 0,15 mgon was achieved.
At the Aggenalm project site Reflectorless Video Tacheometry was used to capture object points distributed on
the slope from a central setup point by means of polar measurements to check for displacements. This is similar
to the conventional setup of a permanent or temporarily tacheometric monitoring system without a geodetic net
measurement. In contrary to a traditional 3D tacheometric system, where prisms are used, in the alpEWAS
approach, only natural targets and no retroreflecting prisms were used. A limitation however is that permanent

measurements may not be possible due to slope retaining snow for extended periods in the Sudelfeld area.
Especially in wintertime long gaps have to be accepted, as the built-in camera is not specified for temperatures
below 0 Celsius.
In Figure 3 the measurement points in the Aggenalm test site are shown. The arrows mark the planar
displacement vectors between fall 2008 and fall 2009 (red/dark) and between fall 2009 and summer 2010
(orange/light). Tacheometric target points without vectors werent measured in all epochs due to single
measurement failures in target detection or range measurement. Displacements measured between 3 and 12 mm
are not significant because of the setup of the measurement pillar position using stable reference points in big
distances. The displacement orientation and absolute value however appear feasible, because they are
comparable to GNSS results.

Figure 3.

Measurement points at the Aggenalm Landslide. The arrows mark the displacement vectors of VTPS
between fall 2008 and summer 2010. Target points without vector havent been measured in all
epochs due to system failures.

The advantage of video tacheometry compared to conventional and approved methods like traditional
tacheometry lies in the variability of possible target structures which can be both artificial (target marks or
reflectors) and natural. Here, depending on the target type, a more or less complex detection algorithm is
necessary, which, as a rule, has to be individually designed for each single case.
The standard processing chain for target detection using VTPS is shown in Figure 4. At the Aggenalm, surface
rocks have been used for natural targeting; but for monitoring purposes in the field also debris scarps, edges of
buildings and the like are possible. The measurement principle is an edge-based matching relying on a teachingphase during the installation of the system. If a real deformation detection is the goal, other methods can be used,
which for example have been evaluated at the Technical University of Vienna using the same prototype
instrument (see also Reiterer et al. 2010). Further applications are described in Thuro et al. (2010).

For every image-based analysis the complexity of the algorithms and therefore the reliability of the result have to
be questioned critically. Especially in industrial image-driven applications, illumination and other environmental
parameters are optimized by default. This is not possible with video tacheometry, especially when working
outdoors. Reliability rates of more than 90 % are achievable in exceptional cases, only. This was the case in the
alpEWAS project, therefore the autonomous video tacheometric measurement system was abandoned in favor of
a semi-automatic, user-controlled measurement process.
Figure 4 shows the flowchart of a target point detection. Measurement images first become homogenized to be
able to use uniform operators and parameters for the deduction of attributes and objects. Final target point
detection itself is mainly done using descriptive rules (geometric, topologic and/or radiometric) or through the
use of matching algorithms. Evaluation of consecutive images imitates integration by time of some normally
distributed measurement deviations in the images and therefore raises the accuracy statistically.

Figure 4.

Flowchart of a target point detection.

Video tacheometry, which was used for displacement monitoring of non-signaled objects in geodesy for the first
time, has a proven functionality, especially under controlled conditions (e.g. indoors). When using the approach
in the field, distinct limitations became apparent mostly because of refraction effects and chaotic short-term
atmospheric turbulences. These influences also affect traditional tacheometric targeting, but are particularly
critical with video tacheometric image processing due to its time-consuming detection algorithms and sensitive
parameterization. Further developments in this field therefore need to be threefold, while results in one branch
will also be input information to the others:

Advances in hardware to go beyond the prototype status. This is primarily the task of the instrument
manufacturers, and is dependent upon economic factors. This makes it a challenge for scientific research
to identify appropriate areas of use as well as to evaluate advantages and limitations.

Further enhancements of algorithms. Reliability and scope of application of usable algorithms need to be
improved and accordingly adopted to the special tasks of video tacheometry. For this, additional fields
of application need to be identified and implemented into pilot projects. An example could be video
tacheometric vibration determination.

2.3

Modeling of atmospheric influences. The main problem of quality derogation of video tacheometric
measurements (e.g. refraction and air flickering) has to be analyzed with a special view to the new
instrument type and, if it proves possible, to be minimized by empirical or modeled corrections. An
according research project currently is planned.

Low-cost Global Navigation Satellite System (GNSS)

The satellite supported monitoring component of the integrative monitoring system at the Aggenalm consists of
an all-weather low cost GNSS measurement system based on simple robust navigation receivers produced for
the mass market. The applied receivers allow a tracking of the American GPS as well as the Russian Glonass
satellites. Such receivers are also capable of monitoring discrete points at the surface of a sliding slope
continuously with sub-centimeter accuracy.
In order to achieve the required accuracy with low cost GNSS, single frequency and carrier phase receivers are
used in combination with a near real time post-processing analysis. This means that over a predefined period of
time, usually 15 minutes, raw data is recorded continuously at different receivers. Using wireless communication
techniques (WLAN) all raw data is transmitted on the fly to a central computing station. As soon as the raw
data from the different locations is available the baseline processing can start immediately. At least one station
should be on stable ground (reference station) and the others are spread on the slope (there are 3 such stations at
the Aggenalm). If there are multiple reference stations available, a geodetic network adjustment based on least
squares is possible. In contrast to real time kinematic applications the entire evaluation of a near real time
approach can be designed and controlled by the operator. In the alpEWAS project it was shown that the near real
time approach achieved accuracies normally obtainable by the use of high-end geodetic receivers.

Figure 5.

Workflow of the Central Control Application (Glabsch et al. 2010a).

In addition to the design and construction of the GNSS sensor nodes, the core component is the software known
as the Central Control Application (CCA). This software is developed by using the graphical programming
language LabView, National Instruments. All steps in the workflow (Figure 5) from configuration and
initialization to continuous data recording and the parallel operating near real time processing, transmission of
results is controlled and executed by the CCA. A modular, prospective design offers the option to integrate a
great diversity of sensors. Only the corresponding program implementation has to be realized. Individual design
of the evaluation is given by the steps 3 and 4. The information obtained from the GNSS nodes (e.g. coordinates,
quality parameters, status information) is stored in a MySQL database and updated every 15 minutes. All data

collected by the geo sensor network is promptly available in a central data sink for an integrated evaluation. For
further technical details of the developed measuring system see Glabsch et al. (2010b) and Glabsch et al.
(2010a).
With the current state of software development on the GNSS sensor nodes, a continuous and robust operation of
the low cost GNSS monitoring component is possible. Presently, malfunctions occur rarely and are often the
result of power shortages. Longer power outages in the 230 V electricity supply system lead to failures in
particular at the central station and the multi-function node in the middle of the slope. Short power black outs
can be bridged by an uninterrupted power supply (ups). In addition to power shortage issues, an autonomous
GNSS system operation is a challenge especially during winter in an Alpine area (e.g. partial complete snow
cover of the antennas and solar panels for days or weeks). Only an appropriate choice of reliable equipment and
the dimensioning of the devices leads to a reliable year-round operation.
Some results of the low cost GNSS monitoring component at the test side Aggenalm are presented in Figure 6.
In each case, the position components (easting Y, northing X, where Y is approximately parallel to the slope
direction) are depicted for the period March 2009 - December 2010 by a moving average filtering based on the
L1-norm estimator. The filter extends over 24 epochs (nodes 2 and 3) and 48 epochs (node 1). Using a time
interval of 15 minutes to acquire the carrier phase measurements the filter length corresponds to 6 or 12 hours.
The choice of filter type and length is a partial consequence of local obstructions at the sensor nodes that lead to
temporarily incorrect processing results. In some cases a successful solution cannot be computed at all. Despite
some technical problems and the resulting data gaps longer-term movement trends from the three sensor nodes
were captured very well. The time series at node 2 and 3 show an acceleration phase occurring during the spring
snow melt (Figure 6).

---- A ---Figure 6.

-B-

-- C --

Variations of the horizontal position of sensor nodes 1 3, time period March 2009 to December
2010, moving average (node 1: 12 h, nodes 2 and 3: 6 h).

Presumably due to a low level of snow cover on the slope in spring 2010 (C in Figure 6) a comparable trend was
not observed in spring 2009 (A in Figure 6). The effects of a heavy rainfall event in June 2009 (B in Figure 6)
are also visible. An enlarged section for the period 15.03.-15.08.2009 shows the impact of the snow melt phase
(A in Figure 6) and the phase of intense rainfall in June 2009 (B in Figure 6) especially for sensor node 2
(Figure 7).

Figure 7.

Sensor node 2, Variations of the horizontal position during 15.03.-15.08.10, moving average (12 h).

The LC GNSS monitoring component based on standard components for sensors, power supply (batteries with
solar panels) and WLAN communication developed in the project at a prototype stage is successfully tested at
the Aggenalm. In the actual configuration a complete GNSS sensor node requires an investment of about 3000.
The developed Central Control Application allows a reliable system operation. Although the achieved accuracies
with filter lengths of about 6 h already permits the possibility to evaluate long-term behavior of the slope, even
with only small changes during the year, the full potential of using simple navigation receivers for surveillance
tasks has not been fully explored. For an improved early warning system a shortening of the filter length is an
essential challenge. Overall, it was possible in this project to develop the use of low cost GNSS sensor
technology for practical application so far that this new measurement technique into practice can be started now.
The results gained demonstrate that a low cost monitoring approach offers an alternative to more precise
tacheometric geodetic surveys or expensive high-end GNSS receivers.
Some more time series of the measuring results will be presented in the paper of Festl et al. (2011) in the
conference proceedings.

3
3.1

Data management and integrative data analysis

Data management

The alpEWAS Control software package (Figure 8) was developed as a modular software package to control the
entire data management within the project. The software has a flexible layout and offers customization of
subroutines for specific measurement systems.
A central component is the open source MySQL database. The link between the geosensor network, (e.g. the
sensors installed on site), and the database is established by so called sensor plugins. In addition to the sensor
control, status monitoring and readout of raw data as well as the communication of information via a
standardized communication protocol is conducted as a first analysis to provide rapid information (1st level
results) from the database for integrative analysis. Various tools permanently monitor the system status in order
to detect the exceedance of critical system parameters as well as failures of individual sensors or subroutines. If
one of these is detected the system administrator is informed immediately, avoiding data loss. The alpEWAS
Live Viewer (Figure 9) is another useful tool for system maintenance, because it informs the user about the
current system status. Furthermore data can be displayed as time series using different filter options and
combinations.

An important interface is the link between the computer onsite and the end user being interested in information
about the current status. Data management centers around server mirroring, ensuring data integration and
redundancy (e.g. master and slave server arrangement). Communication is via a broadband connection to the
project office where an almost unlimited parallel data access from multiple users is possible. Complex and
computationally intensive analysis can be easily distributed to several different computers.

Figure 8.

alpEWAS Control, management and data analysis software.

For data exchange between heterogeneous systems (interoperability) it is possible to query all results by means
of access authorizations via standardized interfaces. Standards such as the worldwide standard for handling

spatial data for an interoperability, as being developed by the Sensor Web Enablement (SWE) initiative of the
Open Geospacial Consortium (OGC) have been taken under consideration.
The main functions of the Live Viewer (Figure 9) are the permanent status information of the systems state and
the possibility to graphically display first results. The user can choose for example which data base (master or
slave) needs to be accessed for live view.

Figure 9.

alpEWAS Live Viewer displaying the status of different sensors as well as an image taken by the
webcam.

Conclusions and perspective

The measurement techniques developed in this project, TDR and LC GNSS, are more or less based on already
existing hardware components, while the video tacheometer hasnt been introduced to the market as yet existing
only as a prototype from Leica Geosystems. New developments in video tacheometry were mostly accomplished
in the fields of adaption of the system layout for undertaking measuring tasks, e.g. slope monitoring and in
signal analysis and sensor control. The efficiency of TDR and low cost GNSS for continuous monitoring was
demonstrated with the development of software components for the data management and complex data analysis

representing decisive steps in practicability. All systems have proven as reliable measurement system at large.
The measurement techniques developed in this project are on the merge of entering the marketplace, where the
range of application isnt restricted to mass movements, but may be extended to other monitoring tasks such as
monitoring of structural elements (Singer et al. 2009a).

References

Festl, J., Singer, J., Thuro, K. (2011). The Aggenalm landslide first findings of the acquired monitoring data. In
International Symposium on Rock Slope Stability in Open Pit Mining and Civil Engineering, Vancouver, Canada 1821 Sept. 2011.
Glabsch, J., Heunecke, O., Schuhbck, S. (2010a). Development and testing of a low cost sensor PDGNSS landslide
monitoring system using the example of the Aggenalm Landslide in the Bavarian Alps. In Altan, Backhaus,
Boccardo & Zlatanova (eds.), Geoinformation for Disaster and Risk Management, Copenhagen, JB GIS, pp. 6370.
Glabsch, J., Heunecke, O., Schuhbck, S. (2010b). berwachung von Rutschhngen mittels Low-Cost GNSS Empfngern
im near Real Time Processing. In Wunderlich (ed.), Ingenieurvermessung 10. Berlin, Wichmann, pp. 275288.
Glabsch, J., Heunecke, O., Schuhbck, S. (2009). Monitoring the Hornbergl landslide using a recently developed low cost
GNSS sensor network. JAG 3: 179192.
Leica Geosystems AG (2007). Leica TPS1200+ Series. XI.07 738600de, Herbrugg, 16 p.
Leica Geosystems AG (2010). Leica Viva TS15. Data sheet, Herbrugg, 1 p.
Reiterer, A., Huber, N., Bauer, A. (2010). Bildbasiertes Erfassen und Matching homologer Punkte mittels Feature-Vektoren
Funktionsweise und Evaluierung beim Einsatz in einem Deformationsmesssystem. In Wunderlich (ed.),
Ingenieurvermessung 10. Berlin, Wichmann, pp. 241254.
Singer, J. (2010). Development of a continuous monitoring system for instable slopes using Time Domain Reflectometry
(TDR). PhD Thesis, Chair for Engineering Geology, Technische Universitt Mnchen, Munich, 189 p.
Singer, J., Grafinger, H., Thuro, K. (2009a). Monitoring the deformation of a temporary top heading invert using time
domain reflectometry. Geomechanics & Tunnelling 2: 238249.
Singer, J., Thuro, K., Festl, J. (2010). Development and testing of a time domain reflectometry (TDR) monitoring system
for subsurface deformations. In Zhao, Labiouse, Dudt & Mathier (eds.), Rock Mechanics in Civil and Environmental
Engineering. London, Taylor & Francis Group, pp. 613616.
Singer, J., Festl, J., Thuro, K. (2009b). Computergesttze Auswertung von Time Domain Reflectometry Messdaten zur
berwachung von Hangbewegungen. In Marschallinger & Wanker (eds.), Geomonitoring, FE-Modellierung,
Sturzprozesse und Massenbewegungen. Heidelberg, Wichmann, pp. 1934.
Singer, J., Schuhbck, S., Wasmeier, P., Thuro, K., Heunecke, O., Wunderlich, T., Glabsch, J., Festl, J. (2009c). Monitoring
the Aggenalm Landslide using economic deformation measurement techniques. Austrian Journal of Earth Sciences
102: 2034.
Thuro, K., Wunderlich, T., Heunecke, O., Singer, J., Schuhbck, S., Wasmeier, P., Glabsch, J., Festl, J. (2009). Low cost 3D
early warning system for alpine instable slopes the Aggenalm Landslide monitoring system. Geomechanics &
Tunnelling 2: 221237.
Thuro, K., Singer, J., Festl, J., Wunderlich, T., Wasmeier, P., Reith, C., Heunecke, O., Glabsch, J., St. Schuhbck (2010).
New landslide monitoring techniques Developments and experiences of the alpEWAS project. JAG 4: 6990.
Topcon Positioning Systems, Inc. (2008). Topcon Imaging Station - Long-Range Scanning, Imaging and Robotic Total
Station. Revision A 7010-0876, Livermore, 4 p.
Trimble Navigation Limited (2010). Trimble VX Spatial Station - Datasheet. 022543-261F, Singapore 4 p.
Wasmeier, P. (2009a). Grundlagen der Deformationsbestimmung mit Messdaten bildgebender Tachymeter. PhD Thesis,
Chair for Geodesy, Technische Universitt Mnchen, Munich, 169 p.
Wasmeier, P. (2009b). Videotachymetrie - Sensorfusion mit Potenzial. AVN: 261267.