Landslides are one of the most destructive natural hazards that threaten human lives and influence the socioeconomic condition of many countries. A landslide is basically identified as slope failure that is caused by gravitational force as the primary factor [1
]. Other triggering factors, such as rainfall, earthquakes, floods or human activity may also influence the occurrence of this geologic event. Because of its specific geologic, morphologic, climatic and tectonic settings, Iran is one of the most landslide-prone areas in the world. By the year 2000, approximately 2600 landslides, which were responsible for 162 deaths, 176 fully destroyed houses and 170 damaged roads, were detected and mapped by experts from various government agencies in the country [2
]. Figure 1
shows a landslide distribution map of Iran. As shown in Figure 1
, most of the landslides are concentrated on the rim of the Alborz and Zagros mountains.
Detection and monitoring of landslides is necessary to decrease and mitigate associated losses. However, this is challenging in slow-moving landslides because of their complicated mechanisms [4
]. Measurement of the surface displacement is also important to predict models of slope failure and to make risk assessments. Remote satellite techniques and in situ
field observations are common methods to investigate landslides and monitor slope instability [5
]. In the present study, we use Interferometric Synthetic Aperture Radar (InSAR) and Global Positioning System (GPS) observations to investigate the slope stability of the Sarcheshmeh landslide, northeast Iran.
The Sarcheshmeh landslide is a northwesterly-directed slope in the province of North Khorasan in northeastern Iran (Figure 2
). A village called Sarcheshmeh with approximately 400 inhabitants is at the foot of this slope, which is 1.6 km long in the east-west direction and 600 m long in the north-south direction, with an approximate area of 100 acres.
The area has an annual precipitation of approximately 300 mm, which occurs during rainy periods from September to April [12
], resulting in a relatively high level of groundwater resulting in springs on the slope (Figure 3b
). The materials involved in sliding are mixtures of soil and rocks in alternating formations of thin-bedded to massive silt clay and limestone; the latter constitutes the majority of the surface geology in northeast Iran (Figure 2b
). The general morphology of the study area is dominated by a mountainous landscape, which includes V-shaped valleys. Preliminary geomorphological analysis suggests that the Sarcheshmeh slope was part of a greater ancient landslide with a depth of approximately 15 m that sustained more than 10 m downward movement [13
]. Numerous scarps and bluffs with height differences ranging from 2 to 30 m are indicative of the ancient activity of this landslide (Figure 3c
). All the current information about the Sarcheshmeh landslide is derived from morphological and surface geology analysis. The current rates of motion are not well known and the creeping mechanism is poorly understood. Because of the placement of the residential area, which is just downstream of the Sarcheshmeh landslide, monitoring is essential to assess the progressive deformation of this potentially dangerous slope and to avoid or mitigate any damage caused by sudden slope failure. Because of the complicated nature of the landslide phenomenon, information about the previous activity of the slope could help us better recognize effective components that are involved in probable slope creeping. Thus, we can benefit from InSAR techniques and extensive archives of SAR data that could assist us in studying and analyzing past mechanisms of slope motion. Additionally, to assess the current rate of deformation on the Sarcheshmeh landslide, we benefitted from the use of GPS geodetic measurements with point-wise reliable results.
2. Data and Methodology
The dataset used in this paper is divided into two categories: SAR satellite images and GPS observations. To evaluate and analyze the past mechanisms of Sarcheshmeh surface deformation, we used the SAR interferometry (InSAR technique). The current rate of deformation was investigated using GPS periodic measurements.
Space-borne Interferometric Synthetic Aperture Radar (InSAR) has developed rapidly over the past 20 years and has proven to be a valuable tool for topographic mapping and surface deformation measurements [15
]. Because of its dense spatial coverage and competitive accuracy, InSAR has now become one of the most preferred geodetic methods to study surface deformation associated with slow-moving landslides [17
]. Advanced InSAR methods such as Persistent Scatterer (PS) [4
] and Small Baseline Subset (SBAS) [24
], which can overcome artifacts arising from atmospheric noise and temporal and spatial baseline decorrelation, enable us to benefit from displacement time-series analysis using a stack of SAR data. Both techniques work better in urban areas because of the high density of man-made objects, which increase the likelihood of finding coherent pixels in a stack of SAR data over time. Because of the non-urban nature of the landslide area in Sarcheshmeh (it is covered partially with grass and other permanent vegetation and used for farming), we used the SBAS technique for surface displacement time-series analysis. The SBAS method uses all possible interferograms with a perpendicular and temporal baseline below a certain threshold to minimize decorrelation effect and topographic errors. The SBAS interferograms are then inverted after phase unwrapping for deformation time series retrieval.
GPS is a high precision positioning technique, which can be used to provide point-wise, semi-continuous maps of crustal deformation that is associated with geophysical phenomena using a variety of static or kinematic techniques [27
]. Due to operating in all weather conditions and independent of direct visibility between stations and 3D precise positioning of specified points, GPS positioning is a beneficial tool in landslide monitoring [28
]. This technique also has limitations, such as choosing the appropriate location for stations so that they have good sky views and are less affected by multipaths [27
]. The choice of the observation mode and processing techniques depends on several practical considerations, such as the accessibility, cost, number of points and required precision [27
Both GPS and InSAR are complementary to one other in that GPS provides detailed ground-truth vector information of displacement for specific localities compared to the extensive spatial coverage provided by InSAR line-of-sight (LOS) observations [31
]. A combination of different ground and space-based geodetic monitoring systems provides promising tools to assess and evaluate geological hazards [31
2.1. InSAR Analysis
For this study we used ASAR images acquired over the study area by the Envisat satellite from the European Space Agency (ESA). Sixteen descending images (Figure 4
) in I2 mode (incidence angle 23°) were obtained for Sarcheshmeh spanning the years 2004–2006 (Table 1
). The data were processed using the time-series approach of Small Baseline Subset (SBAS) in a two-step process: We initially produced a network of small baseline differential interferograms and subsequently did the time-series analysis to derive a one-dimensional Line-of-Sight (LOS) time-series and displacement map of surface deformation on the slope.
The interferometric processing was performed using a standard repeat-pass method implemented in DORIS software, developed by the Delft Institute for Earth-Oriented Space Research (DEOS), Delft University of Technology, which is freely available to the public [34
The topography-related phase correction and geocoding of the interferograms in the interferometric processing were estimated using a three arcsec Digital Elevation Model (DEM) from the Shuttle Radar Topography Mission [35
]. The DEOS precise orbital information of Envisat that was estimated from the DORIS instruments provided by ESA was used during processing to eliminate orbital phase distortion [36
]. We primarily formed 41 interferograms for Sarcheshmeh and then used 34 with a maximum temporal baseline of 311 days and a perpendicular baseline of 494 m for the time-series analysis. The time-series analysis was conducted using the Small Baseline (SBAS) approach that is implemented in StaMPS software [21
]. Figure 5
presents the final network of interferograms used in the SBAS processing.
2.2. GPS Analysis
To evaluate current surface deformation, we used GPS and created a geodetic network of ∼20 stations, including reference and monitoring points in and around the landslide area. Constructed stations are pillars made of reinforced concrete, anchored to a depth of 1 m. The top of the pillars was 0.5 m above the ground. The reference network consists of three pillars constructed on stable localities from the geological perspective, such as bedrock around and outside the landslide area (marked as BM1, BM2, and BM3 in Figure 6
). The monitoring stations were distributed near the cracks and sliding masses inside the landslide area (MP1–MP17 in Figure 6
). The required accuracy for monitoring slow movement phenomena such as creeping landslides should be on the order of centimeters to millimeters [29
]. GPS relative positioning in static mode can achieve millimeter accuracy after considering measuring system error sources [28
]. In this study the GPS field observations were conducted in static mode with four Trimble dual frequency GPS receivers. Four campaigns of GPS field measurements were performed in October 2010, May and August 2011 and April 2012 to derive the contemporary rate of surface deformation on the Sarcheshmeh slope. Figure 6
shows the established network of control and monitoring stations on a 3D topographic map of the area.
GPS observations were performed in static mode, in which monitoring points were positioned by the differential method with respect to reference points. Each monitoring point was observed from at least two reference stations. Additional observations allowed us to control the position of monitoring stations to detect any probable errors. The average session duration for baseline determination between reference points was approximately 7 h, whereas the session duration for the baselines between the monitoring and reference points was 40 min. The collected raw data in each epoch of observation after converting to standard RINEX format were post-processed to achieve geodetic baselines between stations from space baselines between stations and satellites using precise ephemeris [38
]. Figure 7
presents the final network of baselines used for positioning computation using GPS. To obtain compatibility of coordinates with the Iranian National Geodetic Network (INGN), we connected our network to a nearby permanent station from this network. A least-squares optimization method was then applied to adjust the network to determine the unique precise coordinates of stations during each campaign. It should be noted that reference network adjustment computation (baselines between reference stations BM1–BM3) was performed first in each epoch to ensure the stability of reference stations between two sequential campaigns. The displacements of the monitoring stations were computed from the difference between the coordinates of monitoring points determined in each epoch. The standard error analysis using a 95% confidence interval was applied to calculate the relative error ellipse and other statistical parameters that were associated with displacements using a variance-covariance matrix of observed baselines. It is worth nothing that computed formal errors by GPS software are usually over-optimistic [39
]. Therefore a scale factor was computed and applied to the formal GPS errors based on the average of the weighted root-mean-square residual to the linear fit, computed for all stations [40
In this study, InSAR was applied to Envisat data to assess the history of surface deformation on the Sarcheshmeh landslide, and repeated GPS campaign measurements were performed to assess the rate of surface deformation in the area. Although both methods were implemented in different time intervals, both indicate current stability of the Sarcheshmeh slope.
Standard errors analysis in GPS measurement does not show significant displacement on the landslide at 95% confidence interval. As shown in Figure 12
, the cumulative displacement for most of the GPS stations on the landslide during the 2010–2012 time interval falls within the bounds of the estimated error ellipse; the only exceptions were MP.13 and MP.14 which were affected by localized deformation. The large uncertainty of the estimated displacements may arise from the limited number of campaign-mode measurements that we used in this study. The uncertainties of site displacements decrease if more repeat GPS surveys are performed to span longer times or we use continuous GPS instead of campaign-mode measurements [43
For the InSAR investigation, the main limitations of C-band repeat-pass interferometry to reliably evaluate the long-term creep rate over the Sarcheshmeh landslide are atmospheric artifacts in interferograms with short temporal baselines and temporal decorrelation in interferograms with long time intervals (∼1 year); the area is mountainous, partially covered with grass and other permanent vegetation and is used for farming, all of which contribute to the disturbance of the InSAR signal. However, SBAS time-series analysis using Envisat data enables identification of more than 600 coherent pixels that illustrate slope stability in the years 2004–2006 within the error threshold of InSAR observations. The mean deformation rate of the slope with respect to the reference area selected on the residential area downslope is estimated at 1.2 ± 4 mm/year. However, as the main direction of the landslide is towards the NW, any coherent motion related to landslide would have appeared as negative LOS in descending InSAR results. The positive value for the mean deformation and its relatively large uncertainty is an indication of the large impact of residual tropospheric effects on InSAR time-series analysis that compromises the precision of the derived deformation signal using a limited number of SAR images used in this study [44
The Sarcheshmeh slope has been characterized as an ancient creeping hillside. Cluttered topography and the existence of scarps, trenches and deep grooves (Figure 3
) are all geologically indicative of the creeping history of this slide. This area is located on a vast, massive limestone (Figure 2
); thus, creeping can be facilitated by the groundwater level and high precipitation rate among other factors. Surface water can loosen dry soil and make sliding easier on the slippery bedrock. Water penetration also leads to slope overloading and increased slope weight. Therefore, gravity can interact more strongly in downward creeping, especially on weakly consolidated masses. The current stability of the Sarcheshmeh landslide might result from a combined reduction in the ground water levels and precipitation rate. Sarcheshmeh is in an area in Iran with a relatively high level of groundwater and high rate of seasonal precipitation. Multiple springs on and around the slope are indicative of relatively high levels of groundwater.
However because of regional population growth, the expansion of agricultural activity and, subsequently, the irregular use of underground water resources, this area has been subject to tangible reduction in groundwater level over the last few decades. Figure 13a
shows a watershed basin segmentation map of North Khorasan. Different colors in Figure 13b
are indicative of the groundwater level status of the segments. As shown in this figure, the Esfarayen basin is in a critical forbidden area from the groundwater level aspect because of irregular water extraction. The Sarcheshmeh slope, marked by a green dot in Figure 13a
, is in the Esfarayen basin; from 1993 to 2010, the average level of ground water in the basin to which the Sarcheshmeh slope belongs has declined more than 13 m (Figure 13c
). Additionally, the precipitation rate has also decreased by more than 30% during 2004–2007 (Figure 14
). All of these factors might have contributed to the increased resistance of the slope to downward motion.
Despite the recent stability of the Sarcheshmeh slope, our GPS results indicate that human-induced farming activity might trigger sudden motion on this slope. As mentioned in the GPS results section, at two stations, namely MP.13 and MP.14, we recorded significant horizontal movement during our GPS campaigns. Field check-ups revealed that the motion on these stations was due to the development of a new farmland nearby. This farmland was irrigated manually from groundwater resources, facilitating the penetration of water to creep-prone masses. Therefore, saturated weak consolidated soil became heavy and slippery and started to move downward because of gravity. If similar practices become widespread, this accidental event and its consequences, which result from human interference, can be a critical warning regarding the Sarcheshmeh slope. Current geodetic investigation shows that this ancient landslide is primarily dormant in the current period. However, even in the absence of major triggering factors such as earthquakes and heavy rainfall, the landslide can be reactivated by other factors, such as water infiltration from farming. Therefore, preventive measures and subsequent proceedings are still necessary to mitigate and reduce probable losses and causalities. The development of a surface and subsurface drainage system to control and conduct runoff outside the slope, construction of a new residential area in a safe area far from the landslide zone and establishment of an early warning system to provide information concerning critical moments are some executable plans to mitigate any damages.
In this study, time-series analysis of Envisat InSAR data and repeated GPS observations using campaign measurements were employed to assess the slope stability of the Sarcheshmeh ancient landslide in Northeast Iran. No previous studies have examined the past and most recent movement of this landslide, and as a result its creeping mechanism was poorly understood. We were able to perform InSAR time-series analysis using Envisat data and detected more than 600 coherent pixels on the landslide, which showed no statistically significant LOS motion in the years 2004–2006. Horizontal campaign GPS data collected between 2010 and 2012 at 17 stations did not reveal any significant site motions within the estimated error ellipse of GPS results either. We interpret these results as indicating that the Sarcheshmeh landslide is currently inactive. Hydrological analysis suggests that the current dormancy of the landslide might be due to the absence of hydrological triggering factors in the region over the last decade.
Although the InSAR and GPS results in this paper are promising, we find that the uncertainty and precision of the results hamper our ability to detect and analyze the very slow creep signal in this region (<2–3 mm/year). Part of this uncertainty arises from not enough sampling of the slow creep signal in time in both Envisat data with 35-day repeat interval and campaign-mode GPS measurements. For future research, higher temporal and spatial resolution monitoring using observations from new SAR data such as TerraSAR-X (with a resolution of 1–3 m and a repeat time of 11 days) and placing continuously recording GPS sites in the landslide area are recommended. This would allow for more detailed sampling of deformation signal in time and space, in turn mitigating better the atmospheric and noise effects for evaluating slow landslide creep, and improvement of our capability to analyze the magnitude and seasonal dependence of a potential landslide motion in response to triggering conditions.