Evaluating the Performance of GPS Survey Methods for Landslide Monitoring at Hillside Residential

Area: Static vs Rapid Static Othman Z#1, Wan Aziz W. A*2, Anuar A.*3

#Department of Civil Engineering, School of Professional and Continuing Education,University Technology Malaysia International Campus, Jalan Semarak 54100 Kuala Lumpur, Malaysia

1othman08@ic.utm.my *Department of Geomatic, Faculty of Geoinformation and Real Estate, University Technology Malaysia

81310 Skudai. Johor. Malaysia 2waziz@utm.my

3anuara@utm.my

Abstract— The landslide is considered as one of the worst natural disaster that continuously affecting many tropical countries, especially during the monsoon season. Landslides bring destructiveness and various losses to the human living. For the past 25 years, many rainfall induced landslides have occurred throughout our country that strikes the citizens, especially near the hillside areas whereby several properties damaged, human deaths and injured had been reported. Landslide monitoring scheme is therefore very crucial and should be done continuously. Various studies have been conducted to monitor the landslide activity using many approaches, such as classical geotechnical and geodetic surveying method. Each of these approaches posed their own advantages and limitations. This study discusses the effectiveness of satellite tracking data in landslide monitoring. This project area located at Section 5, Wangsa Maju, Kuala Lumpur. The monitoring network consists of four (2) control points namely M01 and G01, and eleven (11) monitoring. Five GPS surveys involving 11 GPS points have been conducted, namely 1st epoch – May 2005, 2nd – November 2005, 3rd epoch – May 2006, 4ft – November 2006 and 5ft – May 2007, respectively separate by using two GPS observation modes such as static (1st,2nd and 3rd epoch) and rapid static (4ft and 5ft epoch) mode. The results of GPS surveys show that the magnitudes of land movements in the study area vary from mm to cm level, depending on the location and the observatiuon period in relation with rainy and dry season. The paper will also discuss the constraints faced by GPS survey method in the landslide prone area environment, which is usually hilly and sloping sharply.

Keywords-component; Evaluating, Performance, GPS, landslide, monitoring, residential area, static, rapid static

I. INTRODUCTION A landslide is considered as one of the worst natural

phenomenon that threat human life and property all over the world, including Malaysia. As one of the developing country, Malaysia has growth with the rapid economic developments over the last decades. Thus, the development of highlands area such housing, highway and golf course construction, intensive

forest logging have resulted in frequent occurrences of landslides. This phenomenon has created a major geo-hazard that claiming life, damaged properties, causing economic disruption, and producing different environmental problems. Most of the landslide tragedies are largely triggered by incidences of heavy rain either a single heavy rainstorm event or successive days of intense rain during the rainy seasons. For the past years, the landslide hazards due to heavy rainfall periods were occurred in many regions in Malaysia as summarized in Table I.

TABLE I.

RAINFALL-INDUCED LANDSLIDE TRAGEDIES IN MALAYSIA

Date of Events Location 11 December 1993 Highlands Towers, Ulu Klang, Selangor 29 August 1996 Pos Dipang, Kampar, Perak

12 January 2000 Kampong Baru Ringlet, Cameron Highlands, Pahang

27 December 2001 Kampong Seri Gunung Pulai, Johor 20 November 2002 Taman Hillview, Ulu Klang, Selangor 31 May 2006 Kampong Pasir, Ulu Klang, Selangor 6 December 2008 Bukit Antarabangsa, Ulu Klang, Selangor

Landslide deformations can be characterized in a variety of ways using different sensor technologies and data acquisition methods. For example, the monitoring can be done using precise survey networks and also by the use of geotechnical tools such as inclinometer, pezometer and accelerometer. Precise surveys include transit traverse survey, triangulation method, leveling survey and total station methods would provide data on the extent of movement on the ground surface. These normally yield information on the X-Y-Z planes, and again can be related to rate of movement. On the other hand, inclinometer data can be used to arrive at the rate of landslide movement and the rate of horizontal movements for embankments and excavations. In landslide monitoring scheme, the landslide has to be represented by discrete measuring points in such a way that the deformations of the

2011 IEEE 7th International Colloquium on Signal Processing and its Applications

453978-1-61284-413-8/11/$26.00 ©2011 IEEE

sliding area can be reconstructed by the measured displacements of those points. Precise survey methods for deformation monitoring provide reasonable accuracy, but resulting in heavy workload, high personnel risk and low efficiency. The use of GPS can considerably improve the situations whereby it can operate in all-weather conditions and intervisibility between stations is not required and also can achieve high accuracy to meet the requirements [5], [6], [10], and [13]. Therefore, an efficient and effective monitoring technique should be established in order to detect the rate of movement, size and the direction of the landslide. Assessment of real situation landslide includes the efforts to monitor land movement continually. [4] had categorized landslide research into three important phases:

a. Define and classify landslide, b. Monitoring activities for landslide, and c. Analysis and movement trend (deformation

modelling).

This paper discusses the observations and analyses of the slope movement using static and rapid-static GPS method. The main objective is to evaluate the performance of static and rapid-static GPS technique in landslide monitoring at hillside residential areas.

II. MONITORING LAND DISPLACEMENTS

Monitoring landslide activity is of paramount importance for landslide studies. Land displacement monitoring in a certain landslide prone area in principle is the monitoring of changes in distances, height differences, angles and/or relative coordinates of the monuments covering the area being studied. Landslide monitoring is generally accomplished by field-based geodetic, space technology (GPS, remote sensing), geotechnical and geophysical techniques. The main investigations are geological structure, surface deformation, ground water and geotechnical. The examples are given in Table II, which is adopted and updated, from [5].

TABLE II. METHODS AND TECHNIQUES FOR MEASURING LANDSLIDE DISPLACEMENT Methods/technique Result Typical

range Typical

precision

Fixed wire extensometer Distance change <10 – 80 m 0.3 mm/30 m

Triangulation Coordinates differences (2D)

< 300 – 1000 m 5 -10 mm

Precise levelling Height change Variable, usually <

50 m 0.2 – 1 mm/km

Electronic Distance Measurement Distance change

Variables. Usually 1 –

14 km

1-5 mm + 1-5 ppm

Aerial photogrammetry

Coordinates differences (3D)

H flight < 500 m 10 cm

GPS survey Coordinates differences (3D) variable 2-5 mm + 1-2

ppm

III. PRINCPLES OF LANSLIDE STUDY USING GPS SURVEY METHOD

Nowadays, Global Positioning System (GPS) had become a useful tool for the positioning of objects. Robustness GPS equipments, its reliability and its ease-of-use are some of the factors why GPS system is popular in survey works. With emerging new GPS technology, many positioning methods and sophisticated software have been developed to collect field data efficiently whether for real-time purposes or post-processing purposes. Thus, GPS technology had became more progressive and has been apply in survey jobs, engineering surveys and other mapping purposes. GPS is beneficial in enabling deformation monitoring..

According to [16], GPS has several advantages over the other types of technology:

i. GPS operates 24-hour in any weather conditions. ii. GPS does not need direct visibility between the base

and monitoring points with the minimal user interaction.

iii. GPS surveying allows acquisition of a large number of high resolution observables at a relatively high speed.

iv. GPS can monitor large areas without a drastic reduction of precision of the measurements for 3-dimensional (3-D) positioning information.

To carry out a research in which GPS technology is used as a tool to provide three-dimensional coordinates for each monitoring point, the network design and the establishment of stable monitoring monuments are vital to ensure the success of such research. It is because the monuments play the important role as the medium in such research that involved the landslide area. In order to explain the deformation of the study area and to obtain more satisfied results, it is better to combine the geotechnical with GPS methods in landslide monitoring – see [8] and [15]. The concept of landslide movement monitoring using GPS technology is explained in Fig 1. In principle, the survey monuments shall be installed in stable ground area as reference stations and at the landslide area as monitoring stations. Often stable ground may be considerable distance away from the area being surveyed for deformation.

Control station

Fig. 1. The Concept of Landslide Monitoring using GPS Technology [4]

Control station Monitoring stations

2011 IEEE 7th International Colloquium on Signal Processing and its Applications

454

IV. GPS SURVEYS IN WANGSA MAJU

The study area is located at the Section 5, Wangsa Maju, Kuala Lumpur, with the latitude, between 3° 11’N and 3° 12’N, and the longitude is between 101° 44’E and 101° 46’E. The existing geological subsurface at the study area was divided into two types of soils: (i) metamorphic stones, and (ii) plutonic stones. The metamorphic and plutonic stones are consisted of the wall-schist and hawthornden-schist, and granite, respectively. Quartzite and phylitte members were also present within this formation. The soils analysis was based on the existing geological investigation reports which have been carried out by local geologist [17].

The GPS survey network in the study area is consisted of two (2) control points established by Department of Surveying and Mapping, Malaysia (DSMM) namely G01 (Gombak) and M01 (Melawati), and eleven (11) monitoring points which were firmly monument on site. Four dual frequency GPS receivers Trimble 4800 have been used to establish the control points, namely M01 and G01, and also monitoring points, marked as WM1, WM2, WM3, WM5, WM9, WM10, WM18, WM21, WM23, WM28 and WM29, respectively. The GPS monitoring network of study is shown in Fig 2.

Fig. 2. The distribution of base stations, control points and monitoring points for landslide area (not to scale)

These GPS points as shown in Figure 2 were selected to enable a reliable detection of landslide displacement signal in the area. At the same time these points should satisfy the criteria for good GPS point, e.g. it is a relatively stable location, has a good sky view and is relatively less affected by multipath [16]. Five GPS surveys involving 11 GPS points have been conducted, namely 1st epoch – May 2005, 2nd – November 2005, 3rd epoch – May 2006, 4ft – November 2006 and 5ft – May 2007, respectively separate by using two GPS observation modes such as static (1st,2nd and 3rd epoch) and rapid static (4ft and 5ft epoch) mode. The distances between the control points with other monitored GPS points are between 2 to 5 km. The GPS surveys at all points were carried out using dual frequency geodetic-type GPS receiver. M01 and G01 were used as the reference (stable) points with known coordinates. GPS observations were conducted with the session lengths of about 1 hour for the static and 15 to 20 minutes for the rapid static. The data were collected with a 15

second interval and elevation mask was set at 15˚ for all points.

V. DATA PROCESSING, RESULTS AND ANALYSIS Results from the landslide study may include several series

of analysis such as GPS network baseline processing, network adjustment (adjusted coordinates), control stations stability test, monitoring stations deformation test and also final analysis for deformation modeling. There are two main phases in GPS data collection. The first phase is to transfer the coordinates of geodetic stations to the reference stations in the research area, and second phase is to collect the coordinates of monitoring stations in the area of interest using these reference stations. In principle, the GPS data processing will involve GPS baseline and GPS network adjustment. The dual frequency GPS data have been processed using the GPS Network Adjustment Program.

The obtained standard deviations of the computed coordinates were typically in the order of several mm for horizontal and vertical component for first, second and third epoch, as shown in Fig 3. This Figure shows that in general standard deviations of the horizontal components for the first, second and third epoch are better than 20 mm, and those of vertical component are better than 50 mm.  

M01

WM1 WM2WM3WM5 WM9

WM10

WM1WM21WM23

WM28 WM29

G01

A01 J01

a) Latitude component

b) Longitude component

2011 IEEE 7th International Colloquium on Signal Processing and its Applications

455

c) Vertical component

Fig. 3. The standard deviations for horizontal and vertical component with G01 and M01 as reference station for epoch 1, epoch 2 and epoch 3

Next, Fig 4 shows the standard deviaitons for the fourth and fifth epoch. This Figure also shows that in general standard deviations of the horizontal components for the fourth are better than 20 mm, and those of vertical component are better than 50 mm. But for fifth epoch the standard deviaition are slightly high at the horizontal component, whereby the value is vary from 22 mm to 30 mm. The values for vertical component for the fifth epoch ar ebetter than 50 mm.

a) Latitude component

b) Longitude component

c) Vertical component

Fig. 4. The standard deviations for horizontal and vertical component with G01 and M01 as reference station for epoch 4 and epoch 5

In this study, landslide displacements are obtained by differencing the coordinates of GPS stations obtain from five consecutive surveys. In this case, the obtained coordinate differences for the horizontal and vertical component between first, second and third epoch are shown in Fig 5. While, the coordinate differences between fourth and fifth epoch are shown in Fig 6.

a) Horizontal component

b) Vertical component

Fig. 5. Coordinates differences betrween 1st , 2nd and 3rd epoch

2011 IEEE 7th International Colloquium on Signal Processing and its Applications

456

a) Horizontal component

b) Vertical component

Fig. 6. Coordinates differences betrween 4ft and 5ft epoch

Figure 5(a) and 5(b) indicate that the difference of coordinate between epoch 1 and epoch 2 is vary between ± 0 to 27 mm. In general, the coordinate differences between these two epochs are considered quite small and can be accepted a good results. But, the coordinate difference between epoch 1 and epoch 3 are quite high at station WM 3 with 35 mm. While the coordinate difference between epoch 2 and epoch 3 for Northing axis is also quite small between ± 0 to 28 mm. Figure 5(b) shows the coordinate differences of vertical axis between epoch 1 and epoch 2 with G01 and M01 as reference station is vary between ± 0 to 24 mm. For epoch 2 and epoch 3, the coordinate differences of vertical coordinate is varying from ± 0 to 41 mm – see Figure 5.88. Next, the coordinate differences between epoch 1 and epoch 3 are between ± 0 to 46 mm.

Next, Figure 6(a) and 6(b) illustrate the coordinate differences for epoch 4 and epoch 5 with G01 and M01 as reference station. Figure 6(b) clearly explained the coordinate difference between epoch 1 and epoch 4, epoch 1 and epoch 5 and epoch 4 and epoch 5, respectively. It is indicate from the figures, the biggest coordinate difference for epoch 1 and epoch 4 is occurred at station WM 3 with 39 mm for Northing axis, 30 mm for Easting axis and 65 mm for vertical axis – see Figure 6(b). These followed by epoch 1 and epoch 5, whereby, the highest coordinate difference is occurred at station WM3 with 41 mm for Northing axis, 31 mm for Easting axis and 66 mm for vertical axis. For the coordinate difference between epoch 4 and epoch 5 are considered quite small for the horizontal and vertical component and can be accepted as good results.

In order to statistically check the significance of the displacement derived by GPS surveys, the static defromation model [15] was performed. The static deformation model is the most basic model used in this work. Its starting point is hypothesis tests detecting coordinate differences and thus movements of control points in between different observation campaigns. These tests are computed for data of two different observation campaigns with the linear Gauss-Markoff model as follows:

⎥⎦

⎤⎢⎣

⎡⎥⎦

⎤⎢⎣

⎡=⎥

⎦

⎤⎢⎣

⎡+⎥

⎦

⎤⎢⎣

⎡

2

1

2

1

2

1

2

1

xx

A00A

vv

ll

⎥⎦

⎤⎢⎣

⎡=

22

112oLL Q0

0QσK (1)

where, l1 and l

2 are the observation vectors at times t

1 and t

2; v

1 and v

2 are the residual vectors at times t

1 and t

2; A

1 and A

2 are

the coefficient matrices at times t1

and t2; x

1 and x

2 are the

vector of unknown parameters at times t1

and t2; Q

11 and Q

22 are the covariance matrices at times t

1 and t

2 and K

LL is the

model’s variance-covariance matrix.

This linear model allows testing linear hypothesis whether control points show statistically any significant movement between times t

1 and t

2 . Hence, the linear H

0 hypothesis:

H0: x

2 - x

1 = [ ] = B X + W = 0 (2)

is tested against.

⎥⎦

⎤⎢⎣

⎡−

2

1

xx

II

H

1: x

2 - x

1 ≠ 0 (3)

The test is computed using the difference vector d and its covariance matrix Q

dd given by

d = x2 - x

1, (4)

Qdd

= Q11

+ Q22

= (A1

T P

1 A

1)

+ + (A

2

T P

2 A

2)

+ (5) which results in the squared sum of residuals θ2 (6) dQdθ dd

T2 +=

where the estimated variance of unit weight as follows

21

22T211

T12

ffvpvvpvs

++

= (7)

The test statistic T can be calculated by

T = hs

dQd2o

ddT +

> F(r,f,1-α) (8)

2011 IEEE 7th International Colloquium on Signal Processing and its Applications

457

The hypothesis Ho is rejected and H1 is true, if T is greater than the upper α-percentage point of the F-distribution with rang r and f degrees of freedom: f i= ni – ui – qi ; r = (f1 – f2) – u ; i=1,2 (9) where n is the number of observations, u is the number of unknown parameters, and q is the number of restrictions. After testing each unknown parameter, the difference vector can be separated into two groups: a fixed group F including those control points where Ho is true, and a moving group M where Ho is false and H1 is true [15]. This results in :

(10) ⎥⎦

⎤⎢⎣

⎡=

M

F

dd

d

(11) ⎥⎦

⎤⎢⎣

⎡==+

MMMF

FMFFdddd PP

PPPQ

so that the difference vector of the moving group is given by dPPdd MF

1MMMM−−=

MF1

MMFMFFFF PPPPP −−= (12)

with the squared sum of residuals 2Mθ

MMMTMFFF

TFdd

TM dPddPddPd +==2θ (13)

In this way, the statistic model can be divided into two models for two groups. Considering only the M group of control points,

iMMM

TM

2Mi )dPd(θ = , i = 1, 2, 3……, n. (14)

the largest value of Equation (14) (15) )max(θ)(θ 2

Mimax2M =

gives an estimate of the maximum deformation that can be expected.

This procedure also results into an iterative working scheme because in the first computation, all control points are included in the datum definition. But after determining the M group of control points, the one with the largest value

is eliminated from the datum definition, then the process repeated, and so on.

max2M )(θ

The results of statistical testing shown in Table III indicate that in general four stations have significant displacements. Table III shows that in the period of May 2005 to May 2006, May 2005 to November 2006, and May 2005 to May 2007 only WM1, WM2, WM5, WM9, WM10, WM28 and WM29 that statistically show no significant displacements, while for WM3, WM18, WM21 and WM23 has a significant displacements. The amounts of displacements for the four

station WM3, WM18, WM21 and WM23 are between 1 cm to 7 cm.

VI. CLOSING REMARKS

Based on the results obtained from five GPS surveys that have been conducted in the landslide area in Wangsa Maju, it can be concluded that GPS survey method is a reliable method for studying and monitoring landslide displacements. Precision level of mm to cm can typically be achieved, although achieving this level of precision is not an easy task to do. Dual frequency geodetic type receivers have been used and it is compulsory used along with a good survey planning, stringent observation and data processing strategy. From this study it can be suggested that in order to conclude the existence of real and significant displacement of GPS stations, the GPS derived computed displacement should be subjected to the static model testing. Moreover, in order to provide physical meaning of GPS derived displacements, the results should be correlated with the geotechnical method such inclinometer to study the area and its surrounding. The GPS derived result should also be integrated with the results obtained by other geodetic monitoring techniques such as leveling and EDM measurements. Finally it should be emphasized that further research is still needed to clarify the real mechanism and pattern of landslide displacements in the study area.

TABLE III. STATIC (θ2-CRITERION) DEFORMATION RESULTS

STATIC MODEL (θ2-CRITERION)

Period May 2005-Nov 2005 (epoch 1 -2) Points 1 2 3 5 9 10 18 21 23 28 29

dx (cm)

STAB

LE

STAB

LE

STAB

LE

STAB

LE

STAB

LE

STAB

LE

STAB

LE

STAB

LE

STAB

LE

STAB

LE

STAB

LE

dy (cm) dz(cm)

Decision �Period May 2005-May 2006 (epoch 1 – 3) Points

STAB

LE

STAB

LE

MO

VED

STAB

LE

STAB

LE

STAB

LE

MO

VED

MO

VED

MO

VED

STAB

LE

STAB

LE

dx (cm) dy (cm) dz(cm)

Decision �Period May 2005-Nov 2006 (epoch 1 -4) Points

STAB

LE

STAB

LE

MO

VED

STAB

LE

STAB

LE

STAB

LE

MO

VED

MO

VED

MO

VED

STAB

LE

STAB

LE

dx (cm) dy (cm) dz(cm)

Decision Period May 2005-May 2007 (epoch 1-5) Points 1 2 3 5 9 10 18 21 23 28 29

dx (cm)

STAB

LE

STAB

LE

MO

VED

STAB

LE

STAB

LE

STAB

LE

MO

VED

MO

VED

MO

VED

STAB

LE

STAB

LE

dy (cm) dz(cm)

Decision

2011 IEEE 7th International Colloquium on Signal Processing and its Applications

458

REFERENCES

[1] Coe, J.A., Godt, J.W., Ellis, W.L., Savage, W.Z., Powers, P.S., Varnes, D.J., Tachker, P. (2000). Seasonal Movement of The Slumgullion Landslide As Determined From GPS ObservationJuly 1998 – July 1999. Open File Report 00-101. U.S Geological Survey.

[2] Cruden D. (1991). A simple definition of a landslide. Bulletin IAEG. 43: 27-29

[3] Kane, W. F. and Beck, T. J. (1996). "An Alternative Monitoring System for Unstable Slopes." Geotechnical News, 14 (3), 24-26

[4] Georg Gassner, Andreas Wieser, Fritz K. Brunner (2002): ‘GPS Software Development For Monitoring of Landslide’. Proc. FIG XXII, Deformation Measurement and Analysis II, Congress Washington, D.C. pp. 12.

[5] Gili J.A.; Corominas J. (2000): Using Global Position System techniques in landslide monitoring, Engineering Geology (special issue), 55:167-192

[6] Guo J J.; Yang Z.; Ding P.; Zhou X Zhu. (2004): Application of GPS on landslide monitoring-A case study of Xiakou landslide, Journal of Global Positioning Systems (2005)

Vol. 4, No. 1-2: 184-191 [7] Hill, C. D. and Sippel, K. D. (2002). Modern Deformation Monitoring: A Multi Sensor Approach. FIG XXII International Congress.

Washington: USA.April 19-26.1–12. [8] Kalkan, Y., Baykal, O., Alkan, R. M., Yanalak, M., and Erden, T. (2002).

Deformation Monitoring with Geodetic and Geotechnical Methods: A case study in Ambarli Region. International Symposium on GIS. Istanbul. Turkey. Sept.23- 26.pg 1 – 12.

[10] Sdao, F. Pascalem, S., and Rutigliano, (2005) : Geomorphological features and monitoring of a large and complex landslide near Avigliano urban area (South Italy), Advances in Geosciences, 2, 97–101, 2005

[11] Tubbs, D.W. (1975). Causes, mechanisms and prediction of landsliding in Seattle: Ph.D. thesis, Univ. Washington. http://www.tubbs.com/disserta/disserta.htm

[12] Varnes, D.J. (1978). Slope movements types and processes. In Landslide : Analysis and Control, Schuster R.L. and Krizek, R.J. eds., Transportation Research Board Special Report 176., National Research Council. National Academy of Science.

[13] Xu, S, W. Cheng, X. Huang, D. Xu (2003): The investigation of landslide monitoring in the Three Gorges Reservoir region by applying GPS, Journal of Hydraulic Engineering (1):114-118.

[14] Wan Aziz W.A, Khamarrul A.R (2003). An Appropriate GPS Technology For Landslide Monitoring At East-West Highway, Perak, Malaysia. Map Asia 2003. Kuala Lumpur.

[15] Yalçınkaya, M and Bayrak, T. (2002). Monitoring the behavior of landslides by GPS measurement: A case study in North Eastern Anatolia (Turkey). Karadeniz Technical University. Turkey. Received 30 April 2003.

[16] Abidin, H, Andres H., Gamal M. Surono and M.Hendrasto (2004). On the Use of GPS Methods for Studying Land Displacements on the Landslide Prone Area.

[17] Zulkarnain, Y. (1997). Geological Structures for Wangsa Maju Areas with related to Slope Stability Study. Undergraduated Thesis at Geological Department Universiti Malaya. Kuala Lumpur.

[18] Wan Abdul Aziz W.M.A., Othman Z.,Halim S., Investigation the Risk of Landslide on the Hillsite Residential Area Using GPS Tracking Data. Map Asia. Jakarta. 2005.

[19] Wan Abdul Aziz W.M.A., Othman Z., Bayrak, T (2004) Landslide Deformation Monitoring And Analysis Using Satellite Tracking Data (Case Study: Residence Area At Wangsa Maju). Paper Presented at International Symposium On Geoinformation, 2004 - 21-23 September 2004, Kuala Lumpur.

[20] Z. Othman, W. Abdul Aziz, T. Bayrak (2009). Landslide Monitoring By Combination Of GPS And Geotechnical Techniques. International Symposium and Exhibition on Geoinformation 2009. August 10-11.

2011 IEEE 7th International Colloquium on Signal Processing and its Applications

459