Introduction

The Thompson River valley south of Ashcroft, British Columbia, Canada (Fig. 1), hosts several landslides concentrated along a 10 km section of the valley. The on-going investigation work at one of these landslides, the Ripley Slide, has been given status of on-going project within UNESCO’s International Programme on Landslides. This corresponds to the continuous resources invested at the site for landslide monitoring and interpretation, focused on new technologies and landslide risk management. The section selected for the study corresponds to an initiative for expanding this in-situ test facility to some of the other landslides in the area. The characteristics of these landslides and their significance have been discussed in several studies examining mechanisms, triggers and deformation behavior (Clague and Evans 2003; Eshraghian et al. 2007; Hendry et al. 2015). Some landslides have failed in a rapid to very rapid motion, although many are considered to be currently inactive or at least moving slow enough not to impact the railway lines that run along the valley (Fig. 1). The few monitoring locations in the valley do not allow for the characterization of landslide activity due to their size and areal coverage. Thus, the interpretation that landslides have been or are inactive is not based on measurements, but rather the absence of notable impact on infrastructure. A typical view of the corridor is presented in Fig. 2.

Fig. 1
figure 1

Location of the study area (a, b). Distribution of landslides (c) with Ashcroft and the Ripley Slide highlighted (Hendry et al. 2015). Also shown are the alignments of Canadian National Railway (CN) and Canadian Pacific Railway (CP). Coordinates in UTM (units in meters for scale purposes). Red dashed line corresponds to the location of a typical cross section of the Goddard Slide (Fig. 4)

Full size image
Fig. 2
figure 2

Photograph of the South Slide and CN54 Slide; view north to Ashcroft

Full size image

Satellite Interferometric Synthetic Aperture Radar (InSAR) was adopted in this study to detect areas along the study area that are moving and provide some insight into their state of activity. Numerous studies have shown that space-borne InSAR can measure deformation over large areas, especially in semi-arid regions with limited snow and vegetation (e.g., Zhang et al. 2016; Shi et al. 2016; Michoud et al. 2016; Novellino et al. 2017). Ongoing extensive monitoring in the Thompson River valley particularly at the Ripley Slide provides an opportunity to compare the displacements interpreted by InSAR to those measured with a differential global positioning system (GPS) and ShapeAccelArrays™ (SAA) installed in boreholes.

This paper provides the geologic context of the area affected by landslides and summarizes the landslide characteristics and mechanisms as interpreted by previous studies. This background is followed by descriptions of the satellite InSAR imaging and processing performed; comparison of InSAR measurements against other instrumentation data at the Ripley Slide; and compared to a building structure known to be stable. The contribution concludes with an interpretation and discussion of the InSAR monitoring results and implications for the activity of landslides in the Thompson River valley. Acronyms are presented as they appear in the text and are also listed at the end of the paper.

Geology of the area and landslide mechanisms

The recent geologic history of the Thompson River valley is best summarized as a series of glaciations punctuated by periods of erosion and valley incision. During multiple Pleistocene glaciations, ice coalesced over the Interior Plateau to form an ice sheet that covered all but the highest peaks. During glacial advance and retreat intervals, the valley infilled with a complex sequence of materials ranging from deposits of poorly sorted sand, gravel and diamicton to rhythmically interbedded silt and clay (Clague and Evans 2003; Huntley and Bobrowsky 2014) (Fig. 3). In postglacial times, the south-flowing Thompson River periodically down-cut through 150 m of these sediments which resulted in the formation of extensive terraces and subsequent landslides that cut into the Pleistocene sediments (Porter et al. 2002). These landslides are translational soil slides that have undergone episodes of retrogression. This is evidenced by a basal through-going shear surface and the presence of steep internal shears. Moreover, the surface expression of several of these landslides is characterized by the presence of multiple graben and horst features. Figure 4 shows a cross-section sketch of the Goddard Slide, typical of the landslides in the area.

Fig. 3
figure 3

Stratigraphic Column typical of the study area (after Eshraghian et al. 2007)

Full size image
Fig. 4
figure 4

Cross-section sketch of the Goddard Slide (after Eshraghian et al. 2008b). See Fig. 1 for location of cross section

Full size image

Rhythmically interbedded, silt and clay deposits near the base of the Pleistocene sequence (Unit 2 in Fig. 3) have been proposed as being responsible for slope instability at many of the landslides within the valley (Clague and Evans 2003; Eshraghian et al. 2007). This unit was deposited during a glacial advance and contains pre-sheared discontinuities that were formed either during periods of glacial advance and overriding or as a result of ancient slope movements (Hendry et al. 2015). Clay materials within Unit 2 have a back calculated residual friction angle between 9° and 16° (Hendry et al. 2015).

Reactivation of these landslides has been attributed to seasonal fluctuations in river level (Eshraghian et al. 2008a, b), which have been shown to correlate with landslide velocity, particularly at the Ripley slide (Hendry et al. 2015). In this regard, Hendry et al. (2015) showed that changes in the river elevation affected the landslide groundwater regime and buttressing effect such that the slope’s safety factor fluctuated between 1.0 and 1.12. High elevations would elevate the groundwater level but the buttressing effect would stabilize the landslide. As the river elevation decreases, a drawdown effect is combined with a loss in buttressing to reduce landslide stability. These variations were associated with periods when the landslide was active (lower river elevations in the Winter months–November through March) and periods when the landslide slows down sometimes coming to a stop (higher river elevations in the Summer and Fall).

Anthropogenic activities contributing to slope instability include the flood irrigation practices introduced to the area with the arrival of western settlers at the turn of the nineteenth century (Tappenden 2014, 2016) and modification of slope geometries during construction and subsequent expansion of the two railway lines running through the valley.

Landslide volumes range between 0.75 million m3 (estimated volume of the Ripley Slide) and 15 million m3 (North Slide) (Macciotta et al. 2016). The estimated rate of movements range from Extremely Slow (less than 16 mm/day) to Very Rapid (up to 50 mm/s) (Tappenden 2016). Many of the landslides have shown periodic increases in velocity, although still classified as Extremely Slow. However, some occurrences of retrogression have been reported with up to Very Rapid velocities (velocity classes according to Hungr et al. 2013).

Only the Ripley Slide is currently being monitored in near real time for deformation and piezometric levels. Other slope instabilities adjacent to the railways are being monitored with periodic slope inclinometer readings and their level of activity is less certain (e.g., the South Slide).

Satellite interferometric synthetic aperture radar (InSAR)

Satellite-based synthetic aperture radar (SAR) transmits microwaves towards the earth’s surface while recording the backscattered signal reflected by the ground. The target area, termed scene, is partitioned into pixels to form a two-dimensional image (Colesanti and Wasowski 2006; Wasowski and Bovenga 2014). The average backscattered signal amplitude and phase within a resolution area (cell) are then recorded for each pixel within the SAR images.

The amplitude data is function of the reflectivity and the geometry of the ground (including vegetation, buildings, targets, etc.) within each cell (Froese et al. 2005), and the phase data are directly related to the distance between the radar antenna and the center of the target. This distance is one dimensional and can only be interpreted along the line of sight (LOS). Since the phase is periodical (limited to 2π) and the number of phase cycles is not recorded between transmission and reception of the signal, the total distance cannot be estimated from the ambiguous phase. This phase ambiguity can be resolved with interferometry by calculating the relative phase difference between two or more SAR images. This is known as differential InSAR (sometimes referred to as DInSAR). Most spaceborne SAR systems operate with wavelengths on the order of a few centimeters (3 to 20 cm), and under ideal conditions ground deformation can be measured over very large areas and with millimeter accuracy (Colesanti et al. 2003; Wasowski and Bovenga 2014). InSAR is sensitive to ground geometry and environmental conditions (Colesanti et al. 2003; Ferretti et al. 2001; Wasowski and Bovenga 2014). In the last few years, advanced processing techniques have been developed to mitigate the effect of these factors. In practice, InSAR is capable of isolating phase changes and interpolating these into surface LOS deformation when consideration is given to geometric, atmospheric, and environmental factors (Ferretti et al. 2007).

InSAR for landslide monitoring

Satellite InSAR has been applied successfully to monitor landslide deformation at local and regional scales. There is a growing body of literature of case studies at regional and local scales where InSAR has been used as a standalone method for the identification of active landslides or in combination with other techniques to improve the understanding of landslide kinematics and to monitor displacement rates. At the regional scale, Novellino et al. (2017) presented a study that covered an area of over 130 km2 in Sicily where active landslides and their velocities were identified. Michoud et al. (2016) successfully identified active landslide areas through a transportation corridor in the Argentinian Andes. Cascini et al. (2010) presented a study that covered an area of about 489 km2 in central-southern Italy. They developed a methodology for landslide identification at the regional scale, and then worked at a local scale to gain insight into landslide kinematics and patterns of deformation. Other notable studies that combined InSAR at regional and local scales include those by Zhang et al. (2016) in the Gansu Province, China; and Necsoiu et al. (2014) in Idaho, USA. Local-scale studies include those by Mateos et al. (2017) who monitored coastal landslides affecting developments in Spain. Frodella et al. (2016) who monitored a rock cliff in northern Italy that threatens an historic town. Schlögel et al. (2016) studied the La Clapière landslide in France, where InSAR monitoring was used to increase understanding of the slope deformation patterns.

Satellite InSAR monitoring in the study area by Kosar et al. (2003) and Froese et al. (2005) covered the period 1997–2004 using coarse resolution images obtained by the European Space Agency (ESA) ERS-1 and ERS-2, and the Canadian Space Agency (CSA) RADARSAT-1. Although insight into landslide displacements in the area was gained, the spatial resolution of newer satellite radars for civilian use has increased and InSAR processing algorithms have improved since these benchmark studies.

Landslide displacements measured through InSAR technologies have also been used for measuring landslide state of activity (Bianchini et al. 2013, Frangioni et al. 2015, Cigna et al. 2012). Activity in these studies is measured or updated following Cruden and Varnes (1996). Adequate measurements of landslide activity in these studies require combinations of InSAR displacement measurements, aerial photographic interpretation (laser scanner imaging provide enhanced information on this regard), history of the site and failure mechanisms, and field investigations; which are then combined through a matrix that aids in updating the landslide state of activity (Cigna et al. 2012). States of activity in this approach correspond to (Cruden and Varnes 1996) “Stabilized” (last moved more than one cycle of seasons ago but causes of movement were removed), “Dormant” (Stabilized but causes of movement remain), “Reactivated” (moving again after being inactive), or “Active” (moving). In this study, InSAR displacement measurements are used to define areas showing deformations. Historical landslide footprints are used to discriminate between Active or Reactivated landslides as well as discard inactive areas based on InSAR displacements, therefore providing some insight into their state of activity. Velocity magnitudes are used for classifying the landslides according to Cruden and Varnes (1996) velocity classes. The results of this work set the basis for resource allocation for further investigation for landslide management and confirm their state of activity and intensity.

Methodology

SAR images used in this study were collected by the Canadian satellite RADARSAT-2 (CSA 2015). The collection of SAR images can be grouped by the beam mode (finer modes cover smaller areas at higher resolutions and wider modes can cover larger extents at reduced resolution), beam positioning, and sensor modes, among others. Image groups are then named depending on these characteristics. Image groups for this study were selected to maximize their availability (scanning period and coverage). East-facing SAR images (named F4N) began in spring 2011, and west-facing SAR images (named U5 and U21) in September 2013. This study includes images collected up to October 2015. RADARSAT-2 InSAR repeat pass interval between acquisitions is 24 days (CSA 2015); however, some longer gaps occurred in our dataset, notably March 15 to May 15, 2014; August 10 to October 15, 2014; and July 15 to September 15, 2015. Gaps in information correspond to periods when the satellite was not scheduled to obtain such measurements. Images were requested at a later date (end of 2015); therefore, continuous imaging was not taken at the time. Fortunately the images available allowed for assessment of areas showing ground deformation. The LOS, duration of monitoring, and spatial resolution of the images collected are summarized in Table 1.

Table 1 Summary of RADARSAT-2 acquisition time and geometry
Full size table

SAR images were processed using Persistent Scatterer Interferometry (PSI). This technique identifies Coherent Targets (CTs) based on low spectral phase diversity and on dominant backscattering coefficient relative to scene average. Those represent locations where the relative phase difference can be reliably estimated through time (Wasowski and Bovenga 2014). Thus, CTs result in high relative coherence, which corresponds to a low variability in the measured phase change (Wasowski and Bovenga 2014). In this regard, threshold values vary for different images and will be a balance between optimal information quality and density. Combined with small InSAR baseline pairs (e.g., all InSAR pairs with perpendicular orbital distance smaller than 200 m), coherent target processing acts to filter atmospheric distortion and other random scatter, greatly reducing their influence on the results. Examples of ideal CTs include buildings, stable rock outcrops, or purpose-built artificial corner reflectors (CR). Figure 5 shows the density of the CTs in the vicinity of the Ripley Slide. The time-series displacement between sequential SAR image acquisitions can be calculated for each CT (Colesanti and Wasowski 2006). The final product comprises a dataset consisting of the geographic coordinates of each CT; a variability of the phase difference; and the cumulative displacement (total displacement relative to the first image) versus time (Colesanti et al. 2003). Further details on the processing steps using this technique are beyond the scope of this paper and can be found in Wasowski and Bovenga (2014), Colesanti and Wasowski (2006), and Colesanti et al. (2003).

Fig. 5
figure 5

U21 coherent targets (CTs) in the vicinity of the Ripley Slide and the pump station. Location of GPS stations and the corner reflectors (GSC) used for comparison are also shown. Coordinates in UTM (units in meters for scale purposes)

Full size image

PSI processing of SAR images acquired with modern satellites achieved accuracies of ±1 mm/year for average displacement rates and ±5 mm for individual displacement measurements (Wasowski and Bovenga 2014; Cloutier et al. 2015). In this study, CT displacements within one standard deviation calculated with RADARSAT-2 data varied approximately +/−1 mm around the mean, (max. variation +/−3 mm by the end of 2015). This suggests adequate precision for landslides showing deformations of about 5 mm or more within the monitoring period, which needs to be considered when assessing areas showing no displacement. The trend of average CT displacements varied about +/−1 mm around expected stable conditions, which is consistent with the accuracy cited from the literature. Location precision for CT displacements is associated with spatial resolution and location error. For fine (F4 N) and ultrafine beam modes (U5 and U21) location error is lower than 15 m. This suggests that CT displacements are averages of areas up to 15 m × 15 m, and therefore smaller localized movements are not considered.

Comparison of measured displacements with other instrumentation

Comparison focused on two locations: the Ripley Slide and a concrete irrigation pumping station on the opposite bank of the river founded on bedrock (Fig. 5). Figure 6 shows the displacements between November 2013 and October 2014 measured by the GPS system (GPS 1, 2 and 3 in Fig. 5) projected into the LOS of U21, displacements resulting from the ground PSI processed U21 SAR data, and those from the corner reflectors installed near the GPS locations (GSC in Fig. 6). Figure 6 presents the U21 displacements as average, median and standard deviation (shown as error bars) of CT points within a 25 m radius from each GPS.

Fig. 6
figure 6

Displacement data for GPS 1 (a) GPS 2 (b) and GPS 3 (c) compared with nearby coherent targets (CTs’) average, median, range within a standard deviation (shown as error bars), and corner reflectors

Full size image

The U21 projected GPS 1 displacements were within one standard deviation of the displacement of neighboring CTs and closely followed the average and median. Displacement of GSC 2 closely followed GPS 1 (Fig. 6a). CTs and corner reflectors in the vicinity of GPS 2 do not follow GPS 2 displacements (Fig. 6b) as closely as for GPS 1, although displacements share a similar trend. Mean CT displacements differ from those of GPS 2 by approximately 10 mm (about 20% of the total displacement of GPS 2). GPS 2 displacements are about 5 mm greater than the 1 standard deviation (10% of total displacement of GPS 2). CT displacement comparisons with GPS 3 showed the greatest variations (Fig. 6c). Mean displacements of nearby CTs captured only 22% (20 mm) of total GPS 3 displacements for the period (90 mm). GPS 3 is located on a steep retaining wall, at a location that has shown differential settlements (Huntley et al. 2016), which is reflected on a larger vertical component of displacement in GPS 3 (58%) when compared to GPS 1 and 2 (20% both). This is a localized phenomena and is likely to divert the displacement of GPS 3 from nearby sections of the landslide, and therefore from nearby CTs and corner reflectors. It is noted that overall displacement trends were compatible between InSAR and GPS 1 and 2. Moreover, InSAR displacement magnitudes were in agreement with GPS 1, and of similar magnitude (however not equal) to GPS 2. Discrepancies between InSAR and GPS 2, as well as those explained for GPS 3 are attributed to the different nature of measurements. GPS measurements correspond to the particular location of the GPS unit, therefore being able to capture localized movements, while InSAR measurements correspond to average displacements within the data’s spatial resolution and location error. In this regard, both measurements are complimentary as InSAR provide displacement trends of the overall landslide areas while GPS measurements provide insight of particular locations and localized trends such as for GPS 3. Interpretation of InSAR results for the other landslides needs to consider that measured displacements are characteristic of area averages rather than potential instabilities of smaller size, however, are adequate for assessing landslide movement as a whole.

As the pump station is a concrete structure founded on bedrock, it is assumed that there are none to negligible displacements and the CTs on this structure should indicate no movement. Figure 7a shows average CT displacements in the vicinity of the pump station (U21). CT range of displacements within one standard deviation is also shown as error bars. The histogram of LOS annual displacement rate is shown in Fig. 7b. CT displacements in Fig. 7 suggest no movement of the pump station. It is notable that within one standard deviation, displacement variation of individual CTs was approximately +/−1 mm around the mean, and reached +/−3 mm by the end of 2015. The trend of average CT displacements varied about +/−1 mm around expected stable conditions.

Fig. 7
figure 7

Average coherent targets’ (CTs’) displacements in the vicinity of the pump station and range of displacements within one standard deviation (error bars) (a), and histogram of Line of Sight (LOS) annual displacement rate (b)

Full size image

Interpreted landslide activity

Figure 8a shows all CT displacement rates in the LOS for U21 in the study area. This figure also shows the footprint of six known areas of instability. CT results were used to (1) highlight zones of concentrated displacements outside of known instability areas (potentially new areas of concern), (2) highlight zones of concentrated displacements within the six known areas of instability, and (3) assess if there are signs of landslide retrogression beyond their known boundaries. In this regard, landslide boundaries had been obtained in previous studies through aerial photograph interpretation and field mapping (Eshraghian et al. 2007; Clague and Evans 2003), and given the location error and spatial resolution of the InSAR dataset, further refining these boundaries was not considered adequate unless signs of substantial retrogression were suspected (i.e., concentrated movement within landslide footprints that extend substantially).

Fig. 8
figure 8

CT displacement rates for the study area (a) and displacement intensity maps for the six zones of concentrated displacement along the LOS of U21 (azimuth 280o, incidence angle 45°). Coordinates in UTM (units in meters for scale purposes). Scale for displacement rates is the same for parts (a) and (b)

Full size image

Figure 8a shows that most CTs outside the footprint of the six known areas of instability did not show significant displacements within the monitoring period (less than 5 mm threshold according to the accuracy discussed earlier). These areas were assessed as “Inactive” (no movement within the monitoring period). The exceptions are (1) area east of the South Slide, however showing a scatter displacements in both directions (positive and negative), suggesting the potential for vegetation or anthropogenic artifacts, (2) an extension beyond the boundary of the Barnard Slide to the north although it is not significant and could correspond to the boundary not being correctly identified previously, and (3) an area south of the South Extension showing similar displacements to this area. These have been assessed as potentially active and further field investigation is targeted to confirm their activity and better define if these are new movements or reactivations. Measured displacements in these areas reach up to 50 mm/year, corresponding to “Very Slow” landslides according to the classification by Cruden and Varnes (1996). Other, localized CT displacements were not considered as they reflect minor volumes and processes relative to the landslides under investigation.

The analysis then focused on the six zones of concentrated displacement highlighted for the study area, all of which are located within, or surrounding historic landslide footprints. Maps of LOS displacement rate were calculated through interpolation of CT LOS displacement results using the Inverse Distance Weighted (IDW) interpolation technique, within each Area of Interest (Areal extent of each interpreted landslide footprint—AOI). Figure 8b shows the six landslide zones and the displacement intensity maps for each along the LOS of U21. Detailed displacement intensity maps in the LOS are shown in Figs. 9 and 10. Figures 11 and 12 present the average and median LOS displacements, as well as histograms of LOS displacements for each zone of concentrated displacement. Interpretation of the behavior of each landslide is based on LOS displacements.

Fig. 9
figure 9

Displacement intensity in the Line of Sight (LOS) for the Ripley Slide (a), South Slide and south extension (b) and Goddard Slide and its toe area (c). Coordinates in UTM (units in meters for scale purposes)

Full size image
Fig. 10
figure 10

Displacement intensity in the Line of Sight (LOS) for the North Slide and its toe area (a), Barnard Slide (b) and Red Hill Slide and its toe area (c). Coordinates in UTM (units in meters for scale purposes)

Full size image
Fig. 11
figure 11

Average, median and range of displacements within one standard deviation (error bars) along the Line of Sight (LOS) (left). Histograms of LOS displacements (right) for a Ripley Slide, b south extension of the South Slide, and c Goddard Slide toe area

Full size image
Fig. 12
figure 12

Average, median and range of displacements within one standard deviation (error bars) along the Line of Sight (LOS) (left). Histograms of LOS displacements (right) for a North Slide toe area, b Barnard Slide, and c Red Hill Slide toe area

Full size image

Different landslide orientations imply different orientation of landslide displacement vectors that cannot be compared directly. LOS displacements are projections of real landslide displacement vectors. Average landslide displacements are expected to be sub-parallel to the downslope direction. This assumption is based on the displacement patters at the Ripley Slide, which show surface displacements to be sub-parallel to the downslope direction (Hendry et al. 2015), and that the landslides in the valley show similar kinematics and behavior. Given these characteristics, downslope displacement was considered an adequate and stronger metric for comparing different landslide rates as opposed to LOS displacement. Caution is required, however, due to the uncertainty associated with this assumption, and comparisons should be taken as general assessments for prioritizing further investigation of the landslides in the area. In this context, comparison of landslide displacements in the downslope direction play a substantial input for efficient allocation of resources for further monitoring and investigation.

Average and maximum values of landslide LOS displacements were converted to equivalent downslope displacements (D slope ) through an inverse projection (Eqs. 1 and 2) (Colesanti and Wasowski 2006). In Eq. 1, the dot represents the vector dot product; \( {\overline{U}}_{LOS} \) is the unit vector in the direction of the LOS; \( {\overline{U}}_{slope} \) is the unit vector in the downslope direction; and S represents a scalar factor (secant of the angle between \( {\overline{U}}_{LOS} \) and \( {\overline{U}}_{slope} \)). In Eq. 2, D LOS is the displacement measured in the LOS.

$$ \mathrm{S}=\left(1/{\overline{U}}_{slope}\bullet {\overline{U}}_{LOS}\right) $$
(1)
$$ {D}_{slope}=S\times {D}_{LOS} $$
(2)

The inverse of S represents the percentage of movement captured (P C ) by each viewing geometry as per Eq. 3. The higher P C is indicative of higher precision of CT deformation measurements (Colesanti and Wasowski 2006).

$$ {P}_C=\raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$S$}\right.\times 100\% $$
(3)

In this study, P C values and CT densities were most favorable for the U21 geometry and therefore the results are presented and discussed in detail. Table 2 shows the summary geometries of each displacement zone and the Cartesian components of the slope vector \( {\overline{U}}_{slope} \) associated with each AOI and U21 \( {\overline{U}}_{LOS} \) components used to calculate displacement components in the downslope direction for the average and maximum LOS displacements.

Table 2 Geometric summary of concentrated displacement zones
Full size table

Ripley slide

The Ripley Slide is the most densely instrumented and active AOI within the study area and is ideally oriented for U21 LOS which captures 82% of ground movements (percent capture). CT density was 1505 points/km2. The displacement intensity map (Fig. 9a), LOS displacement time series, and histogram (Fig. 11)a indicate the average ground movement is − 39 mm/year (with a standard deviation of 12 mm/year). This zone is concentrated within the center of the sliding mass. The fastest displacements were detected upslope from the railway tracks. Median values of CT displacements suggest seasonality, with faster rates over the winter months (November through March) and slower movement in the summer and fall. This seasonal variation is consistent with previous observations (Hendry et al. 2015; Macciotta et al. 2014; Schafer et al. 2015). Overall, average and maximum downslope displacement rates at the Ripley Slide have been calculated at 49 and 77 mm/year, respectively. This maximum rate is also consistent with these previous studies. The Ripley Slide has not undergone previous rapid movements. Given the reported displacement rates, this landslide is at an Active state with a Very Slow velocity (according to the classification by Cruden and Varnes 1996).

South slide

The displacement intensity map (Fig. 9b), LOS time series, and displacement histogram of the South Slide (Fig. 11b) indicate no displacement of the main body, with the exception of localized zones attributed to local instabilities or surface processes, for example deflation and gulley erosion. Because the causes of movement are still present, this area was assessed to be Dormant. The area south of the historical limits of the South Slide was found to be the largest active movement zone detected within the study area. The scale of this active area, approximately 181, 000 m2, was not known to the railway companies, nor has it been previously documented. For the South Slide and the south extension, the U21 dataset captured 69% of ground displacements and CT density was 1304 points/km2.

This active area does contain three slope inclinometers (SI) located within the proximity of the railway tracks (near the toe of the slope) (Fig. 13). Inclinometers BGC04-01 and BGC04-12 remain active to this day. BGC04-05 was active until 2010 when it is assumed that the displacement made the casing unreadable (Newcomen and Pritchard 2009; Roberts et al. 2015). The location of these inclinometers is shown in Fig. 13 which also shows the displacement intensity map of this part of the landslide and the CTs located in the vicinity of each inclinometer (U21).

Fig. 13
figure 13

Location of slope inclinometers (SIs) in the south extension of the South Slide and displacement intensity map (a), detail of coherent targets (CTs) near SIs BGC04-01 (b), BGC04-12 (c) and BGC04-05 (d)—U21 Line of Sight (LOS). Coordinates in UTM (units in meters for scale purposes)

Full size image

Measurement frequency of the SI was once or twice per year and the BGC04-05 monitoring period did not overlap with the InSAR monitoring period. The SI show annual horizontal displacement rates of 3.2, 16.0 and 3.5 mm/year for BGC04-01, BGC04-05, and BGC04-12, respectively, and suggest extremely slow displacement rates at the southern landslide boundaries, increasing towards the north.

Slope inclinometers measure the horizontal component of landslide movement. These were projected to the horizontal component of the LOS direction (azimuth 280o) and compared against CT displacements within a 25 m radius from the SI locations (Table 3). It is noted that LOS horizontal component would not capture all horizontal displacement of the landslide; however, this allowed for a high level evaluation of the consistency between SI and SAR measurements. CT displacements were consistently larger than SI measurements, but of similar magnitude, with SI measurements within 1 standard deviation of CT displacement values except for BGC04-01.

Table 3 Slope inclinometer (SI) displacements at the south extension of the South Slide and average coherent targets (CT) displacements in their vicinity. Displacements are projected to the Line of Sight (LOS) horizontal component and normalized to annual rates for consistent comparison
Full size table

SAR analyses identified the overall extent of deformation of the south extension of the South Slide. Displacements at the south extension appear to be largest at the center of the mass and near the crest. CT displacement median values also show seasonality (Fig. 11b). Average LOS displacement rates at the south extension are − 17 mm/year (with a standard deviation of 8 mm/year) (Fig. 11b). Average and maximum downslope displacement rates were calculated as 21 and 60 mm/year, respectively. This corresponds to a Very Slow landslide according to Cruden and Varnes (1996). Because this area has not undergone any previous rapid motion, it is believed that this is the first displacement period for the overall South Extension, therefore this area’s state of activity is Active as opposed to Reactivated according to Cruden and Varnes (1996).

Goddard slide

InSAR coverage of ground displacement was lower at the Goddard Slide when compared to the other AOI, with a percent capture of 42% and CT density of 493 points/km2. The displacement intensity map is shown in Fig. 9c. LOS time series and displacement histogram are shown in Fig. 11c. These indicate that the active zone is concentrated at the toe of the landslide and a zone near the crest. Localized zones of deformation are present within the main slide body between the toe and the crest. Time series data from the toe (median values) suggest there may be a weak seasonal variation at this site, with observed displacement rates being most elevated from November through February. Average LOS displacement rates at the toe of the landslide are − 14 mm/year (with a standard deviation of 10 mm/year). Average and maximum downslope displacement rates were calculated as 31 and 120 mm/year. This would correspond to a Reactivated landslide with a Very Slow Velocity at the toe area, CT displacements seem to indicate the body of the landslide above the toe is currently Dormant or showing Extremely Slow velocity (less than 16 mm/year).

North slide

InSAR displacements at the North Slide were concentrated at the toe of the landslide. Most of the landslide body showed no displacement with only localized zones with increased values. At the toe of the landslide, the U21 dataset captures 59% of downslope displacements with a CT density of 492 points/km2. The displacement intensity map (Fig. 10a), LOS time series, and displacement histogram (Fig. 12a) capture the behavior of this landslide. The active toe is in close proximity to a section of railway track. Average LOS displacement rates at the toe of the landslide are − 14 mm/year (with a standard deviation of 13 mm/year). Average and maximum downslope displacements are 25 and 81 mm/year. Deformation does not show any clear seasonality. Similar to the Goddard Slide, this would correspond to a Reactivated landslide with a Very Slow Velocity at the toe area, CT displacements seem to indicate the body of the landslide above the toe is currently Dormant or showing Extremely Slow velocity (less than 16 mm/year).

Barnard slide

InSAR displacements at the Barnard Slide are relatively slow compared to other AOI. This landslide has a good CT coverage, with a percent capture of 85%. CT density is 703 points/km2. This value is lower than other areas with high percent capture, which can be attributed to increased vegetation cover in the upper terrace. The displacement intensity map is shown in Fig. 10b. LOS time series and displacement histogram are shown in Fig. 12b. These suggest that the Barnard Slide is most active between Spring and early Fall (CT median values), which is opposite to that observed at other AOI. Although further monitoring is required to validate this observation, this could suggest different mechanisms are influencing the behavior of this landslide when compared to the known behavior of others (e.g., Ripley Slide). Average LOS displacement rates are − 11 mm/year (with a standard deviation of 8 mm/year). Average and maximum downslope displacements are 10 and 50 mm/year. This corresponds to landslide velocities classified as Extremely Slow to Very Slow (Cruden and Varnes 1996). An extension of the displacement area was identified beyond the boundary of the Barnard Slide to the north although it is not significant and could correspond to the boundary not being correctly identified previously. The landslide is assessed to be at a Reactivated state.

Red Hill Slide

The Red Hill Slide is an historic failure whose footprint contains a smaller area of active displacement at the toe of the landslide. U21 coverage is low at this slide (percent capture of 37%) however there is a CT density of 1109 points/km2. The displacement intensity map is shown in Fig. 10c; the LOS displacement time series and histogram in Fig. 12c. The Red Hill Slide downslope movement would be expected to follow an eastward direction (towards the river, see Fig. 8). All other landslides discussed in this study have a downslope northeast to southeast direction (towards the river). LOS displacements are measured relative to the satellite position, and displacements with opposite directions will show different signs. Therefore, for a same satellite dataset (U21 in Figs. 8, 9, and 10), displacements at the Red Hill Slide have positive values while the other landslide movements are shown as negative values (Fig. 10c and 12c).

CT median displacement values show strong seasonality. Periods of increased displacement occur in the late fall and winter months and decrease significantly from mid-spring through to summer. The seasonality observed here is similar to that at the Ripley Slide (located in the proximity of the Red Hill Slide—Fig. 8). This suggests similar mechanisms could be acting at both AOIs (e.g., river buttressing effects at the toe and rapid draw down mechanisms). Extensive river erosion and sloughing of the riverbank has also been observed at the toe of the landslide, which is likely contributing to the observed displacement behavior. Average LOS displacement rates are 19 mm/year (with a standard deviation of 14 mm/year). Average and maximum downslope displacements are 37 and 89 mm/year, similar to those at the Ripley Slide. The landslide is assessed to be at a Reactivated state showing Very Slow velocities.

Landslide displacement comparison

A summary of displacements is presented in Table 4. LOS displacements (average values) have been used to evaluate the individual landslides. Displacements are measured along the LOS between the satellite and the AOI, therefore their value depend on the orientation of the landslide’s displacement vectors relative to the LOS. In turn, this is a function of the slope orientation and the characteristic of the landslide kinematics. As previously mentioned, landslides in the area have shown similar behavior and kinematics. In this regard, Colesanti and Wasowski (2006) present an approach to compare the displacement of scenes with different geometry with respect to the LOS by projecting the LOS displacements to the downslope direction. Although downslope projections would provide approximations of real landslide displacements, the methodology provides a more consistent value for comparing landslides with different geometries with respect to LOS. Downslope displacements were calculated for each landslide average and maximum LOS displacement measurements. Figure 14 shows the time series of average downslope displacements for the six AOIs. Table 4 and Fig. 14 suggest that on average, the Ripley Slide is the most active of the AOIs, with up to twice the LOS displacement rate than some AOIs (North Slide toe), and significantly larger than others (Barnard Slide). However, for most AOIs the maximum downslope displacements are equal to or larger than those at the Ripley Slide, which suggest the potential for localized, faster movements. Differences in average landslide displacements appear to be consistent (Fig. 14), so can be taken as a good indicator of the relative differences in the displacement rates of these areas.

Table 4 InSAR displacements—U21 dataset
Full size table
Fig. 14
figure 14

Time series of average downslope displacements for the six areas of concentrated displacement

Full size image

Summary and conclusions

The landslides along the Thompson River Valley, south of Ashcroft in British Columbia, have a long history of activity. Several previous studies have described their geologic context, mechanisms, triggers and instability factors. These landslides are important since they can potentially impact the two largest Canadian railways operating in the valley and may contribute to significant environmental and economic consequences. Although proactive track maintenance and slope monitoring are in place, an update on the overall landslide activity is required for further allocation of hazard management resources.

This study applied InSAR to measure slope displacements in a key area of southern British Columbia. The findings presented are based on RADARSAT-2 satellite U21 geometry. The contribution details the methodology followed for post processing and comparison of InSAR data, in addition to providing insight into the landslide state of activity in the area and the velocity classes for the areas showing ground displacements within the monitoring period. InSAR technology and processing algorithms have been previously applied successfully to monitor landslide phenomena. Further validation of the technique for this particular site involved direct comparison with ground instrumentation, and evaluating InSAR displacements and variability of a structure known to be stable (concrete structure on a rock outcrop) in the study area. InSAR provided similar displacement values to those observed with traditional displacement monitoring instruments, and displacement variability at the stable location was about +/−1 mm.

The analysis identified six areas of concentrated displacements within the study area, all within or adjacent to the footprints of past landslide occurrences. Average line of sight (LOS) displacement rates range between 11 mm/year (Barnard Slide) and 39 mm/year (Ripley Slide). These correspond to projected downslope displacement rates between 10 mm/year and 49 mm/year, respectively. However, the maximum downslope displacement rate of 120 mm/year was found at the toe of the Goddard Slide. Maximum downslope displacements were between 73 and 89 mm/year for the other landslides, except for the Barnard Slide with a maximum downslope displacement of 50 mm/year. In this regard, most landslides display Very Slow velocities (according to the classification of Cruden and Varnes 1996), with some areas showing Extremely Slow velocities (Sections of the Barnard Slide). Use of InSAR has allowed us to identify landslides showing an Active state of activity (Ripley Slide, South Extension—see Fig. 8) a Reactivated state (Goddard Toe, Barnard Slide, North Slide Toe, Red Hill Slide—see Fig. 8), a Dormant state or showing Extremely Slow velocities (South Slide, Goddard Slide main body, North Slide main body—Fig. 8). Most notably, the large area of movement south of the historical limits of the South Slide has been recognized as Active and is now targeted for further investigations to confirm its state of activity and evaluate the associated hazards and risks.

Most of the Thompson River valley landslides exhibit seasonal changes in the rate of displacement (median displacements particularly at the Ripley Slide, Red Hill Slide toe, Goddard Slide toe, and south of the South Slide). These typically show increased displacements during the late fall and winter months versus relative stability during the spring and summer. This suggests that factors that affect the stability of these landslides are common across the study area, and that these landslides are marginally stable. However, small changes in river elevation and possibly geometry may lead to movement acceleration. These results will be the basis for further deployment of ground instrumentation for detailed monitoring and evaluation of landslides the area.