There are numerous methods used nowadays to monitor landslide movements. Of these methods, Global Navigation Satellite System (GNSS) and Interferometric Synthetic Aperture Radar (InSAR) are the ones that are most commonly used. In this study, the amounts of movements acquired via these two methods were compared and relations between them were analysed. The Koyulhisar landslide region was selected as the field of study. In this study, 10 Envisat images of the region taken between 2006 and 2008 were evaluated using Persistent Scatterers Interferometric Synthetic Aperture Radar (PSInSAR) technique and annual velocity values at the direction of line of slight at PS points were obtained for the region of interest. The velocity values were then obtained from PSInSAR results and compared with those obtained from six periods of GNSS measurements that were performed between April 2007 and November 2008 on Koyulhisar Landslide area after which the relationship between the two was analysed. Two different movement models from GNSS and PSInSAR results were fit to the landslide region. The velocity values estimated from these movement models for the region were compared and correlation between them was determined. As a conclusion, a high correlation of r D 0.84 was determined between the models obtained from nine GNSS points, except one point at the city centre, and PSInSAR.
Trang 1Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=tgnh20
Geomatics, Natural Hazards and Risk
ISSN: 1947-5705 (Print) 1947-5713 (Online) Journal homepage: http://www.tandfonline.com/loi/tgnh20
Comparing the results of PSInSAR and GNSS on slow motion landslides, Koyulhisar, Turkey
Kemal Ozgur Hastaoglu
To cite this article: Kemal Ozgur Hastaoglu (2016) Comparing the results of PSInSAR and
GNSS on slow motion landslides, Koyulhisar, Turkey, Geomatics, Natural Hazards and Risk, 7:2, 786-803, DOI: 10.1080/19475705.2014.978822
To link to this article: http://dx.doi.org/10.1080/19475705.2014.978822
© 2014 Taylor & Francis
Published online: 14 Nov 2014
Submit your article to this journal
Article views: 110
View related articles
View Crossmark data
Trang 2Comparing the results of PSInSAR and GNSS on slow motion
landslides, Koyulhisar, Turkey
KEMAL OZGUR HASTAOGLU*
Department of Geomatics Engineering, Faculty of Engineering,
Cumhuriyet University, Sivas 58140, Turkey
(Received 11 April 2014; accepted 15 October 2014)
There are numerous methods used nowadays to monitor landslide movements Of these methods, Global Navigation Satellite System (GNSS) and Interferometric Synthetic Aperture Radar (InSAR) are the ones that are most commonly used In this study, the amounts of movements acquired via these two methods were compared and relations between them were analysed The Koyulhisar landslide region was selected as the field of study In this study, 10 Envisat images of the region taken between 2006 and 2008 were evaluated using Persistent Scatterers Interferometric Synthetic Aperture Radar (PSInSAR) technique and annual velocity values at the direction of line of slight at PS points were obtained for the region of interest The velocity values were then obtained from PSInSAR results and compared with those obtained from six periods of GNSS measurements that were performed between April 2007 and November 2008 on Koyulhisar Landslide area after which the relationship between the two was analysed Two different movement models from GNSS and PSInSAR results were fit to the landslide region The velocity values estimated from these movement models for the region were compared and correlation between them was determined As a conclusion, a high correlation of r D 0.84 was determined between the models obtained from nine GNSS points, except one point at the city centre, and PSInSAR
1 Introduction
Landslides occupy an important part of natural disasters Landslides, particularly occurring near settlement areas, cause loss of life and property Therefore, monitor-ing landslide movements is very important The Global Navigation Satellite System (GNSS) and Interferometric Synthetic Aperture Radar (InSAR) methods are most widely used in monitoring landslides The GNSS method has been frequently used to monitor landslides particularly in the last 20 years (Gili et al.2000; Malet et al.2002; Coe et al.2003; Hastaoglu & Sanli2011) Similarly, the InSAR method has been fre-quently used in monitoring landslides in the last 20 years as well (Fruneau et al
1996; Singhroy et al.1998; Rott et al.1999; Crosetto et al.2005; Motagh et al.2013) While monitoring landslide by GNSS method produces three-dimensional (3D) and high-precision information, the method itself is demanding and time-consuming Moreover, deformation information obtained by GNSS is point-based and presents
no information on deformations in wider areas If one desires to obtain information about wider areas by GNSS method, measurements are needed to be done at
*Email:khastaoglu@cumhuriyet.edu.tr
Ó 2014 Taylor & Francis
Vol 7, No 2, 786803, http://dx.doi.org/10.1080/19475705.2014.978822
Trang 3numerous GNSS points This has a negative effect in terms of cost and time Further-more, no observation could be performed in monitoring the movements in volcanic and landslide areas since access to such regions is very difficult The areal deforma-tion can be obtained by using interpoladeforma-tion methods from Global Posideforma-tioning System (GPS) results Therefore, it is considered that in studies performed for monitoring landslides, the use of methods such as Persistent Scatterers Interferometric Synthetic Aperture Radar (PSInSAR) that can provide areal movement data along with the high-precision, point-based GNSS measurements is beneficial InSAR methods have been commonly used in recent years to monitor areal deformations in landslide areas
InSAR has the potential to detect ground surface motion phenomena with the accuracy of a small fraction of the radar wavelength on large areas with high spatial resolution (Pratti et al.2010) Permanent Scatterers technique called PSI (Persistent Scatterers Interferometric) method was developed by Feretti et al (2001) Then, the method was improved by many researchers (Berardino et al 2002; Hooper et al
2004; Kampes2005)
The PS (Permanent Scatterers) approach is based on a few basic observations There are ground targets that maintain a coherent reflectivity to the radar in time even when observed from different looking angles (the PS) The interferometric phase in corre-spondence of these targets is not randomized by temporal and geometric decorrelation phenomena (Pratti et al.2010) Decorrelation is caused by contributions from all scat-terers within a resolution cell summing differently, due to relative movement of the scatterers and/or a change in the looking direction of the radar platform “If, however, one scatterer returns significantly more energy than other scatterers within the cell, the decorrelation phase is much reduced This is the principle behind a ‘persistent scatterer’ (PS) pixel, also referred to as a ‘permanent scatterer’”(Hooper et al.2012) PSI technique has been used by many researchers in monitoring the landslides (Cole-santi & Andwasowski 2006; Farina et al 2006; Meisina et al 2006; Herrera et al
2009; Notti et al 2010; Righini et al 2010; Liu et al 2013) The PSInSAR is an advanced technique in comparison with conventional InSAR technique It has many advantages to overcome the problems of decorrelation for generating a time series of phase changes without atmospheric and DEM (Digital Elevation Model) residual effects, so the PSInSAR method is preferred
There are studies in recent years in which GNSS and PSInSAR results were used together (Peyret et al.2008; Yin et al.2010; Catal~ao et al.2011; Cigna et al.2012; Akbarimehr et al.2013; Zhu et al.2014) 3D movement amounts can be determined
by GNSS, whereas one-dimensional (1D) movement amounts can be found at the line of slight (LOS) using the SAR method Therefore, 3D GNSS results are con-verted into 1D results in the LOS direction in order to explain the situation using the two methods together
In this study, GNSS results obtained from Koyulhisar landslide area were evalu-ated together with the PSInSAR results Hastaoglu (2013) conducted six-period GNSS observations at 10 GNSS points in the landslide area between April 2007 and November 2008 and obtained 3D annual velocity values for the points The current study evaluates 10 descending Envisat images taken between 2006 and 2008 by PSIn-SAR method The main reason that the number of images is limited by 10 was that there are only 10 images belonging to the study area in ESA’s archive where GNSS observations were made between 2007 and 2008 Annual velocity values belonging to the PSI points at the LOS direction were obtained as a result of the evaluation
Trang 4Then, a section profile was specified which intersects the landslide area through northsouth direction The PSI points were determined on this section Both of the results obtained from two methods were compared through the study field by trans-forming the 3D velocity GNSS values on LOS direction For the comparison pro-cess, two separate movement models (equations (4) and (5)) were fit for the velocity values on the LOS direction obtained from GNSS and PS methods LOS velocity values at every 50 m were then calculated using the models through the section repre-senting the landslide field Finally, the correlation between these velocity values cal-culated was determined
2 Field of Study
Koyulhisar is 180 km away from Sivas, Turkey Since the study area lies upon the Northern Anatolian Fault Zone (NAFZ), which is an active fault, the rock masses in the region contain discontinuities and are usually seen to be cracked and crushed Depending on the steep topography in the region, there are many old and new land-slides The direction of motion of these landslides usually threatens residential areas (Sendir & Yilmaz2002)
Koyulhisar district centre is located at a region on the NAFZ which is one of the most important seismic belts in the region Landslides constitute a great risk due to both the lithological properties of the rocks existing in the region and the morphology shaped by intense active faulting For this purpose, various observations were made in the region for scientific and technical purposes in different periods (Toprak1998; Sendir & Yilmaz
2002).Figure 1presents the general geological conditions of Koyulhisar region
Koyulhisar and the landslide regions are located very close to NAFZ which is one
of the biggest active earthquake belts One of the most important features of active earthquake zones are specific land forms Along with land forms the mass move-ments resulting in changes in land forms such as landslide, rock fall, and soil fluction, are natural events frequently occurring on such active belts
Old landslide masses are seen in areas close to Koyulhisar district centre and its surrounding where Eocene aged clayey formations, Lower Miocene aged clayey and gypsum formations and Plio-Quaternary aged sediments are observed Most of these landslides have a circular-failure mechanism Koyulhisar district centre is located on
a former landslide which has a circular-failure mechanism The former landslide mass has continued its activity over time However, this activity is not mass type but local landslides occurring in the main mass (Tatar et al.2000)
As a result of the analysis of the landslides that have occurred on the study area, it was acquired that rainfall and the flora plays an important role Especially, the cracks and fissures filled with water due to high rainfall between the winter and spring sea-sons of 1998 and 2000 before landslides caused the strength of clay fillings to be reduced, thus contributing to the movement along failure zones In addition, this has also contributed to the increase in unit weight of soil burden covering the rocks and thus increasing the extra burden on the side slopes Dense forests over the side slopes
in the study field have slowed down the flow of water through the slope and eased the water to seep into the soil material As a result, forces causing the landslide (failure)
on the slope have increased (Sendir & Yilmaz2002)
It was reported in the previous studies (Sendir & Yilmaz2002) that the potential slope instabilities in the study field were generally towards the south (S, SWSE)
Trang 5These are the dominant slope directions in Koyulhisar district and its surrounding It was observed that the former and new landslides determined within the study field between ¸Sıhlar Fault in the south and Dumanlıca Fault in the north were explained
by a complex landslide system which is composed of a combination of many land-slides through Dumanlica and ¸Sıhlar Fault from north to south
The material accumulated in the region after the landslides occurred between 1998 and 2000 is located in G€onenli stream that is located on the east of Aklan district which is on the north of Koyulhisar The width of this slide reaches up to 2 km The ground water level in the region is very high and small lakes were formed in the slid-ing mass It is highly probable that slidslid-ing mass can move again in a season with high rainfall since it would increase the contact of the mass with water and thus the mass would become saturated
Figure 1 Geologic map of the Koyulhisar Landslide area
Trang 63 Data-sets and methods
3.1 GNSS velocities
A GNSS study was conducted in the region by Hastaoglu (2013) and 3D annual veloc-ities belonging to 10 GNSS points in the region were obtained The velocveloc-ities obtained are given intable 1 In order to determine the GNSS velocities, six-period GPS cam-paigns were conducted between April 2007 and November 2008, covering about 1.5 years Each GPS campaign was carried out on three consecutive days Observing session duration for static GPS measurements was about 12 h The data-sampling rate and elevation cut-off angle were set to 30 and 15 s, respectively For each day and each rover point, a position was computed for each 12-h session in ITRF 2005 using BERNESE 5.0 relative static baseline processing strategies (Beutler et al.2005) The accuracy of velocities obtained from GNSS measurements were investigated in detail
by Hastaoglu and Sanli2011 The deformation rates from the GNSS time series were extracted using kinematic Kalman filtering method, and estimated GNSS coordinates from BERNESE were the output for kinematic Kalman filtering method The loca-tions of GNSS points are shown infigure 2 The annual velocities of GNSS points are given intable 1 Statistical tests of the expanded model were conducted, and it was decided that the model consisting of velocities were significant (table 1) The velocity values were divided by its root-mean-square error, and test values were computed Sta-tistical tests were conducted as mentioned previously and results are shown in the deci-sion column of table 1 If parameters have significantly changed in the kinematic model, a “p
” sign is given intable 1 Otherwise, a “¡” sign is given
3.2 SAR data-set and interferometric processing
PS algorithms operate on a time series of interferograms all formed with respect to a single “master” SAR image It is ultimately the level of decorrelation noise that defines whether pixels are PS pixels or not, but an initial selection of candidate PS pixels can be made using various proxies, the most common of which is amplitude dispersion (Ferretti et al.2001; Hooper et al.2012)
Table 1 Movement parameters determined with a kinematic model between April 2007 and
September 2008 (Hastaoglu2013)
) 0.5(¡) 0.9(¡)
) 0.1(¡)
) 0.2(¡) 1.1(¡)
) 5.0(p
) 0.2(¡)
) 0.3(¡)
Trang 7In this study, the data-set of SAR images acquired by the Envisat satellites (descending orbits track 78) was collected from ESA archive For the PSI analyses,
10 archived raw Envisat images were used These images have descending geometry acquired between October 2006 and December 2008 SAR data dated 22 July 2007 was chosen as master and nine interferograms were calculated from this master image In this study, SAR images are gathered in raw format and converted to SLC (Single Look Complex) images with ROI_PAC public software For the interfero-gram processing is applied with the public domain Delft object-oriented radar inter-ferometric software (DORIS) Interferograms are produced by using Delft precise orbits The topographic effect is reduced by an external DEM as 3 arcsecond SRTM data which has 90 m resolution Geocoding is referred with World Geodetic System
1984 (WGS84) reference system In this study, Stanford Method for Persistent Scat-terers (StaMPS) approach was applied for the monitoring of Koyulhisar landslide
PS process was applied with nine interferograms with the pairs which were limited with a 494-m perpendicular baseline and 525 days’ temporal baseline For the PSI analyses, amplitude dispersion index was chosen as 0.3 A list of perpendicular base-lines and temporal basebase-lines is shown intable 2
The annual velocity values at the LOS direction were obtained for PS points given
infigure 3by PS process Analysingfigure 3, it is seen that there is a subsidence on the former landslide mass and there is an uplift on city centre which is located on fur-ther south In order to observe the deformation on PS points in detail, a section was specified through the landslide area (figure 3) and the velocity values of PS points along this section were analysed
Figure 2 Processing of GNSS baselines: CMYK and IKYK are fixed (Hastaoglu and Sanli
2011)
Trang 8Figure 3 Annual velocities of PS points.
Table 2 List of Envisat Asar data used in the PSI approach
Trang 9As shown infigure 3, apart from the given profile zone, there is a subsidence up to
¡12 mm located at the forested area in the north of study area and also there is an uplift up to 11 mm located to the west of centre, presenting the maximum subsidence and uplift values As you see in figure 4, in these areas standard deviations are the highest of all and these areas are out of the profile zones which is chosen as study area The reason of high standard deviations may be the number of interferograms used in the PS process Unfortunately, there are only 10 images belonging to the study area in ESA’s archive where GNSS observations were made between 2007 and
2008 In the literature, it is suggested to use at least 15 images for PSI processes It is also mentioned in Colesanti and Andwasowski (2006) and Notti et al (2010) Hooper et al (2007) showed that using 12 interferograms are usually sufficient when using StaMPS Analysing figure 4, it is seen that standard deviations are quite low
Figure 4 Standard deviation of PS points
Trang 10for the profile zone in study area This situation shows us that the velocity values of
PS points on profile zones are accurate values On the other hand, the time series of some PS points in the area including KH07 and KH10 are investigated (figure 5) As
a result of this investigation, it is showed that closer PS points usually show similar deformation trend As the deformation trend is similar, it can be concluded that there
is no atmospheric or unwrapping error in the processes Finally, although 10 inter-ferograms are used for PS processing, especially the results of PS process in the land-slide area and the city centre are quite safe
4 Comparing the results of GNSS and PSInSAR
In order to compare the GNSS and PSInSAR results obtained, a section line com-posed of three parts (northsouth direction) was specified through the top to the toe
of the landslide area (figure 6) Then, the velocities associated with GNSS and PS points along this section line PS points which are not located 200 m away from the section line which is used to determine the points within the study field were excluded from the data-set The section graph of the velocities associated with the PS points within the landslide area is given infigure 6
GNSS velocities obtained are three-dimensional In order to compare them to PSInSAR results, 3D GNSS velocities were transformed into 1D velocities in the LOS direction by using the formulae given in equation (1) Then, LOS values associ-ated with GNSS points given infigure 7were calculated through the section line spec-ified on northsouth direction within the landslide area
The 3D (in the east, north, and vertical (up) directions) orthogonal components
of the surface displacement of a point on the Earth’s surface is stated as
D ¼ ðdx; dy; dzÞT The formula of projection of the surface displacement vector D to the line of sight can be written as
s ¼ ð¡ cos ahsinu sinahsinu cos uÞT (2) where dLOS, s, and D and denote the LOS displacement, the satellite unit vector, and the surface displacement vector, respectively For a detailed description of the satel-lite unit vector and its parameters, see equation (1) (Arıkan et al.2009) I use master acquisition parameters for heading, ah and u incidence angle, of an Envisat I2 descending pass Using those common GNSS points of measurements in equations (1) and (2), velocities in the radar LOS were computed The results of GNSS are illus-trated infigure 7
Figure 7shows the velocities associated with GNSS and PS points along the sec-tion line As the figure is analysed, it is seen that the velocity value of the GNSS point
at 4700 m of the section is way too different than the others This is the GNSS point denoted by KH07 When the GNSS velocities of the point KH07 is analysed, a hori-zontal movement different than the other points in the landslide area is observed In Hastaoglu’s (2013) study, it was stated that this movement might be due to a local movement independent from the landslide movement in the region They also indi-cated the instantaneous change of water level in the well loindi-cated within the Police Department and emphasized that the velocities found for the point KH07 showed that it is a local movement different than the main landslide movement On the other