Abstract— Rain fall induced landslides are a common cause of damages to life and property in the Western Ghats region in south India. Work have been in progress to develop a monitoring system to predict the landslides to reduce the loss of human life. We have developed and deployed a Wireless Sensor Network to monitor rainfall induced landslide, in Munnar, South India. A successful landslide warning was issued in June 2009 using this system. This paper discusses the enhancement of the existing system by incorporating a Wireless Geophone Network to locate the initiation of landslide, and the direction and velocity of motion of the slide. A nested geophone methodology and triangulation method was designed to collect and analyze the relevant signals. A novel signal processing algorithm was developed to analyze the geophone data and automatically detect the landslide signal. A feedback method used to reduce the traffic congestion in the network is also detailed here. The design and developed system was tested and validated, in the landslide laboratory set up at our university, for which results are shown in this paper.
Keywords- Geophone Signal Processing, Landslide detection, Wireless Geophone.
I. INTRODUCTION Landslides are a common disaster in the western ghat
region of India. In the past they have caused wide damages to human life and property. In order to reduce the damages monitoring and prediction of landslides are being developed.
We have developed and deployed a Wireless Sensor Network (WSN) to monitor and predict rainfall induced landslide. The deployment was done in Munnar, one of the Landslide prone areas in the Western Ghats, India. The wireless technology has provided solutions to monitor otherwise inaccessible and remote areas. A successful warning was issued during the monsoon 2009 using the system. The warning issued facilitated evacuation and disaster management in the area.
The system consists of more than 50 sensors including the pore pressure transducers, dielectric moisture sensors, strain gauges, rain gauges, tiltmeters, and geophones. The
This work is a part of Wireless Sensor Network for Real-time Landslide Monitoring project, funded by Department of Information Technology (DIT), India.
dielectric moisture sensor detects the level of water saturation in the soil. Strain gauges sense any deformation movement. Tiltmeters are used to validate the strain gauge measurement. A geophone was used to record the ground vibrations [1]. Different combination of these sensors are interfaced to and deployed in several Deep Earth Probes (DEPs).
This paper focuses on the geophone sensor, the design of the interfacing circuits and related signal processing algorithm. Geophones can produce readings that, once analyzed, can locate the beginning position of movement and thus predict the direction of motion. This wireless geophone network will be incorporated with the existing system, wireless network for landslide detection, and early warnings will be real-time streamed to the internet.
The remainder of the paper is organized as detailed. Section II describes related work in landslide detection and signal processing. The design is elaborated on in section III. Section IV describes the signal processing algorithm. Finally, section V concludes with brief description of future work.
II. RELATED WORK Krohn et al. explains how to place the geophone properly
in the ground, accentuating ground coupling to provide worthwhile data [6]. Baule et al. has developed a system to detect the ground vibrations using geophones [7]. The detected ground vibrations are processed to produce audio output. Shinji discusses how to distinguish between ground vibrations and noise thus clearing the data received by the geophone [4]. Arattano et al. uses geophones to locate where the landslide is about to initiate [5]. The landslide initiation point is located by analyzing the measured distances between the sensor columns of a WSN. This was a simulation study. Mario et al. used geophone data to indicate the direction of movement of the landslide [3]. Zan et al. uses geophones to provide early warnings for landslides [2]. This system used Mobile phones and local alarms to issue the warnings.
A WSN for detecting rainfall induced landslides has been fully operational, three years prior to this paper. The deployed system is capable of issuing local alarms as well as online streaming through internet. This paper describes the incorporation of a Wireless Geophone Network (WGN)
2010 International Conference on Computer Applications and Industrial Electronics (ICCAIE 2010), December 5-7, 2010, Kuala Lumpur, Malaysia
to the existing system. The WGN has validated in a medium sized laboratoryUniversity. Results of the tests are presented
III. WIRELESS GEOPHONE NET
The developed WGN is capable ofexisting WSN system by automaticallyinitiation point of the landslide. The systemof predicting direction and velocity the lanA geophone is used to capture the vibrationthe landslide event. It is a transducervibrations and converts them into an electelectrical signals from the geophone aretransmitted through the wireless network. end the data is analyzed and processed.
Not all vibrations obtained from the gassociated to a landslide. So the data frorequires noise isolation and analysis. This by proper signal conditioning and processinan important role is also played by the apprand placement of geophone.
A. Geophone Selection The ground vibrations are 3-Dimension
longitudinal segment called the P waves, tracalled the O waves, and the surface wavwaves. Geophones respond only to wavesaxis [8]. So we may need three geoporthogonal to each other to capture all the our pilot deployment, we begun with the geophone, buried deep in a bore hole. Svibrations, created during the horizontal mthe geophone was placed perpendicular toearth. However this one dimensional geprovide enough information. So a 3C geophindividual geophones oriented orthogonal tis selected to be used in the new WGN. Thelps to determine the direction of movemen
Incorporating all these ideas it was decid3C geophones in our existing monitoring sindividual geophones are designed to be coand reference [7] claims that this is effectivto sense very low frequency infrasound grou
Figure 1: Flow diagram showing geophone dat
Figure 2: DEP with nested geophonegeophone is placed in three sensitive lay
been tested and y set up at our d in this paper.
TWORK f enhancing the y predicting the m is also capable ndslide will take. ns induced during r which senses trical signal. The e enhanced and At the receiving
geophone can be m the geophone can be achieved
ng. Furthermore, ropriate selection
nal waves with a ansverse segment ves called the S vertical to their phones oriented three waves. For one dimensional o to capture the
movement of soil, o the surface of
eophone did not hone having three to each other and
The 3C geophone nt of soil layers. ded to implement ystem. The three
onnected in series ve when needing und vibrations.
B. Signal Conditioning The geophone data is preprocessedincrease the SNR. As shown in Figof the geophone data involves ampshifting, A/D conversion.
Landslide induced vibrations athe range of 200 mV. Active amimpedance, in the range of 10-10measure these small range vibratidescribes one of the advantages of that; it ensures a negligible load on a reduction in dynamic resolutionfiltered using an active low pass minimized by using a high cutoff fr
The filter is accompanied by a act as an AC voltage sources; as contains both positive and negativsensor motes work only with positgeophone readings need to be repreThis task is performed by a level sdata is then fed to the WSN interfato a digital signal and further proce
ta processing
e assembly. Note the 3C yers.
d to reduce the noise, and gure 1 the preprocessing
plification, filtering, level
re very small, usually in mplifier with high input 00 M�, was chosen to ions. The reference [9] f high input impedance is
the geophone and avoids n. The amplified data is filter. The data loss is equency filter. level shifter. Geophones such the geophone data
ve values. Since wireless tive values, the negative sented as positive values. shifter. The level shifted
ace, where it is converted essed to be sent over the
70
wireless network. The data is relayed throcluster to a base station. The geophone dthrough a satellite link and finally received Management Server (UMS) to be further an
C. Geophone Placement and Integration Sensors Landslide event detection requires sp
DEP in each location. The design and spatigeophysical sensors on the DEP are determfactors such as: the number of soil layers, laproperties, the presence of impermeable laheight, bed rock location, depth of thedeploying the DEP, and the specific deprequired for each geophysical sensor.
The soil is made up of impermeablelayers. The impermeable layers of soil gather, creating a perched water table, whsoil particles [13]. So theoretically speimpermeable layers which lead to slope inthe ones which requires close monitoringbore hole interbedded permeable and impcan be witnessed. A minimum of one geopin each of the impermeable layers. Figure 2design with nested geophone assembly.
Bore holes are drilled and the DEgeophones are placed in it. The nested geocould help in identifying the area, depth awhich causes the instability. These nassembly are deployed in such a way thattriangulation technique as shown in fig 3geophones decreases from toe of the hill hill. It is the toe region of the hill, wheinitiates so the more set of nested geophonethe toe when compared to the crown and thof the hill.
These nested geophones are connected tonode on the top of the DEP. This will samtransmit the data to its higher layer network
Figure 3: Triangulation method for Deploying the nGeophone
ough the network data is then sent at our University
nalyzed.
with the Other
ecific design of ial distribution of
mined by different ayer structure and ayers, water table e bore hole for ployment method
e and permeable allow water to
hich loosens the eaking it is the nstability and are g. In each of the permeable layers hone is deployed 2 shows the new
EPs with nested ophone assembly and specific layer nested geophone t they perform a . The density of to crown of the
ere the landslide es are deployed at he middle region
o wireless sensor mple, collect, and .
IV. WIRELESS GEOPHONE SThe main aim of the signal proce
identify the landslide induced vrelevance and produce an alert. Ingoals it is necessary to classifyspurious noises or landslide inducednoise removal technique is used. Tfiltering followed by a threshold bclassified signals are then correlatedneighboring DEPs. These steps mafalse proof automatic detection. Onas landslide signals, they are analydirection the slide will take and to lof the slide. A. Digital filtering
Geophones indiscriminately vibrations from traffic, human walketc. In this scenario all vibrations otland movement are not relevant aLandslide induced ground vibratiofrequency, below 5 Hz. A person could induce a vibration around 20digital bandpass filter was usedvibrations from the signal. This is any vibrations other than landslidfurther interpretation.
B. Threshold based Classification
The filtered signal is now readymethod of analysis is to check the As suggested in Mario et. al. [3usually of high amplitude and their voltage is around 200mV, so ththreshold. Signals above this thrclassified as landslide signals. landslide signals undergo further pr C. Vibration detection for confirmin
The purpose of this step is tsystem, to further classify which noises or landslide induced vibratiocorrelation analysis method introduc
If a geophone registers a landindicate that there is a chance that also register the same landslide sigtime delay in the reception of signthis context. The time delay is equfor vibration to traverse the disgeophones. The time delay (�t) shown in Eq (1).
�t = D/v
Where D is the distance between velocity of the wave in soil [13]. T
nested
SIGNAL PROCESSING essing is to automatically ibrations, confirm their n order to achieve these y which vibrations are d vibrations. A three step he first level is a regular based classification. The d with the signal from the ake sure that we have a ce the signals are labeled
yzed further to predict the localize the starting point
register all sorts of king and land movement ther than those caused by and require nullification. ons are generally of low
walking on the surface 0 Hz and higher [10]. A d to remove irrelevant a very crucial step since de signal could corrupt
y for analyzing. The first data against a threshold.
3] landslide signals are approximate lowest peak his was chosen as the reshold, of 200mV, are
Only these classified ocessing.
ng the slope instability to provide a false-proof
vibrations are spurious ons. This is done with a ced by Terzis et al [11]. dslide signal this would
a nearby geophone will gnal. But there will be a nals at the geophones in
uivalent to the time taken tance between the two could be determined as
(1)
the DEPs, and v is the The velocity of sound in
71
Figure 4: Data from geophone without filteringdata is clogged with noise.
soil is pre calculated. If the neighboring gereport a landslide signal within this calculatinitial landslide signal will be classified as s D. Localization of landslide initiation
The calculated time delay between geopas previously discussed, can also be usedstarting point of a landslide and also predictland will slide. To sense land movement,(contained in a DEP) is strategically placedthe landslide prone area. When the land malso moves. This movement can be calculathe initial distance between each DEPs. If Dbetween DEPs before the landslide initiatiodistance after then d2 could be found usshown in Eq 2.
D1/D2=�t1/ �t2 �t1, �t2 are the time delays before and afinitiation. E. Effective data collection and aggregattraffic congestion
According to the Nyquist Criteria the nrate of the geophone should be greater resonant frequency. This means that geophigh sampling rate. When monitoring increase the load on the network. To combathe landslide signals are distinguished froma threshold algorithm. This allows only the to be sent over the network therefore reduciOur system is based on three levels of warnlandslide warning is given when moisturecross a threshold value. A second level landgiven when the pore pressure sensor rthreshold value. A third level landslide wwhen geophone sensor readings registelandslide signal. It is the geophone resultslandslide is definitely about to occur.
g at source. The
eophone does not ed time delay the
spurious.
phones reception, d to localize the t the direction the , each geophone
d in the ground of moves the DEPs ated by knowing D1 is the distance on and D2 is the sing the relation
(2)
fter the landslide
tion for reduced
normal sampling than twice the
phones require a 24/7 this could at this extra load,
m the noise, using landslide signals ng the data load.
ning. A first level e sensor readings dslide warning is readings cross a warning is given er a confirmed s that confirms a
All minute variations in geophonto landslide prediction; therefore been developed. Adaptive samplinsampling rate of the geophone warning. The sampling rate of thewith each higher level of warningwarning is given the sampling rate of the sampling rate at a first level w
V. TESTING AND VALIDATIOLABORATORY SE
The design has been tested in asetup at the University. The mediulong by one meter wide by 0.5 mecapable of holding 0.6 meters of sof mechanically simulating the diflandslide prone area. It is also able tof seepage and rainfall rates. The ssetup along with the associated senWater is then added in the form until the slope fails. The setup helthe various sensors in before field din better understand the nature of was done with a slope angle of 35rate possible.
The lab setup was used to estaplacement, for the geophones, in a first few tests, the geophones werefrom toe to crown. Straight line plaresulted in observing that the initiafrom the toe. Therefore a decision geophones at the toe than at the cknown as triangulation and is furthTests were carried out to establish active low pass filter in the interfnoise. Figure 4 shows the results of pass filter and Figure 5 shows the use. Comparing figure 4 & 5, it is aclogged with noise, and this makes
Figure 5: Data from geophone with aacquired from the test done in our medipeaks represent the landslide induced vib
ne data could be relevant adaptive sampling has
ng involves varying the dependent on level of
e geophone will increase g. When a third level of will be almost thrice that
warning stage.
ON IN THE LANDSLIDE ET UP a medium scale landslide um lab setup is 2 meters eters tall rectangular box soil. The setup is capable fferent slope angles of a to simulate various levels oil is packed into the lab nsors for the experiment. of rainfall and seepage,
lps in testing, calibrating deployment. It also helps landslides. WGN testing 50 at the lowest seepage
ablish the most effective landslide scenario. In the placed in a straight line cement of the geophones tion of the slide happens was made to place more
crown. This technique is her outlined in Figure 3. the necessity of using an facing circuit to remove not using the active low active low pass filter in
apparent that data can get s it difficult to recognize
an active filter. The data is um landslide lab set up. The
brations.
72
Figure 6: Data from geophone without filtering at source. The data is clogged with noise.
the peaks. These peaks are the actual landslide signal. Figure 6 displays the output from the vibration detection algorithm. The circled peaks are the peaks which correspond to the landslide vibrations.
VI. CONCLUSION AND FUTURE WORK We have deployed a Wireless Sensor Network (WSN)
for predicting landslide. To enhance the capabilities of this deployment we proposed a nested wireless geophone network and associated signal processing algorithm. The design was formulated from various recent works in the area and our experience from the pilot deployment. In our pilot deployment we have done a prototype using one dimensional geophone. In the proposed design we use a 3 axis geophone. Such a system is more effective in localizing the slip location and detecting the direction of movement of the soil layers. Since geophone is a self-excited component, it helps in reducing the power constraint on the design of the system. The signal processing algorithm also takes care of reducing the load on the WSN by selected sampling of geophone data. The rate of sampling differs for each level in our three tier warning system. These proposed design changes will help to build a more effective warning system for landslides.
Our pilot deployment was for landslide prediction. In future the system will be extended to monitor avalanche, debris flow and earthquakes. When it comes to debris flow geophones could be used to calculate the mean flow velocity. Furthermore we will be developing a system to image the layers of earth using the geophone. The application of such a system could help localize the slip and advanced prediction of events.
ACKNOWLEDGMENT I would like to express the gratitude for the motivation
and research solutions provided by Dr Sri Mata Amritanandamayi Devi, The Chancellor, Amrita University. I also like to thank Ms. Erica Thapasya S. Fernandez for her valuable contribution to this work.
REFERENCES
[1] M.V. Ramesh, “Real-time Wireless Sensor Network for Landslide Detection”, Proceedings of The Third International Conference on Sensor Technologies and Applications, SENSORCOMM 2009, IEEE, Greece, June 18-23, 2009, pp. 405-409
[2] L. Zan, G. Latini, E. Piscina, G. Polloni, and P. Baldelli, “Landslides early warning monitoring system”, Geoscience and remote sensing Symposium, IGARSS 2002, IEEE International, Vol 1, pp.188-190, 2002.
[3] L. R. Mario, G. D. Saccorotti, G. T. Stefano, C. Giovanni , D. P. Edoardo, “Seismic Signals Associated with Landslides and with a Tsunami at Stromboli Volcano, Italy”, Bulletin of the Seismological Society of America, Vol 94, pp: 1850-1867, 2004
[4] S. Shinji, “A new system to validate Algorithms for early Earth quake detection”, Railway Technology Avalanche, Vol 12, pp:72, March 2006
[5] M. Arattano and L. Marchi, “Systems and sensors for debris flow monitoring and warning”, Sensors, Vol. 8, pp. 2436-2452, 2008
[6] C. E. Krohn, “Geophone ground coupling”, Journal of GeoPhysics, Vol. 49, pp. 722-731, 1984.
[7] H. Baule and M. Borgers, “Acoustic ground vibration detector”, The journal of Acoustical Society of America, Vol. 87, 1990, pp. 923
[8] D. Lawton and M. Bertram, “Field test of 3-Component geophones”, Canadian journal of Exploration Geophysics, Vol. 29, 1993, pp:119-129
[9] A. L. Hagedoorn, E. J. Kruithof and P. W. Maxwell, “A practical set of guidelines for geophone element testing and evaluation”, First Break, Vol. 6, Oct. 1988, pp: 325-331.
[10] H. Baule and M. Borgers, “Acoustic ground vibration detector”, US Patent file, Patent no: 4849947, Jul. 1989
[11] A. Terzis, A. Anandarajah, K. More, and I. J. Wang, “Slip surface localization in wireless sensor networks for landslide prediction”, Proceedings of the Fifth International Conference on Information Processing in Sensor Networks 2006, IPSN 2006, Nashville, USA, pp. 109-116, 2006.
[12] M. Hurlimann, D. Rickenmann, C. Graf, “Field and monitoring data of debris-flow events in the Swiss Alps”, Canadian Geotechnical Journal, Vol 40, pp: 161-175, Feb 2003.
[13] A.T. Kunnath, M.V. Ramesh, “Integrating Geophone Network to Real-Time Wireless Sensor Network System for Landslide Detection”, Proceedings of The fourth International Conference on Sensor Technologies and Applications, SENSORDEVICES 2010, IEEE
