Introduction

Landslide is a natural geohazard that is induced by the action of gravity (Kavzoglu et al. 2014), crashes down without notice, and travels several kilometres. It can cause significant destruction to lives and property globally (Kundu et al. 2013; Di Martire et al. 2016). The two main classifications of landslides are the nature of the movement and the type of materials involved. The movement includes rockfall, internal debris flow, and mudflow that takes place (Schulz et al. 2017) while the materials involved are solid rock or unconsolidated sediments (Furuichi et al. 2018) that are weaker sediments, which have not been compressed considerably by overlying materials (Tappenden 2014). A landslide can either be triggered by natural phenomena such as the action of gravity, snow melting, and continuous rainfall (Mostafa 2012) or by human-made activities such as extraction of oil from the ground, lateral slope cutting for buildings, and movements of human activities (Virk et al. 2019). When it happens, the implications reflect the number of events, like, the presence of people, constructed roads and buildings, and agricultural farming in the environment (Chen et al. 2017).

Table 1 presents some of the past landslide events in different areas of Penang Island between 2016 and 2019. The causes of the hazard, magnitude of the disaster, and district areas they fall in are also shown in the table. The Penang flash-flood in 2017 was a rare catastrophe that the residents will never forget. The island was collapsed by the worst flash floods ever and heavy rainfall that triggered the extraordinary magnitude of the landslide cases. The affected areas include Penang Hills, Paya Terubong, Batu Ferringhi, Tanjung Bungah, etc. These landslides became active and prompted the slope in these regions to shift. Figure 1 shows a few disasters of the landslide occurrences due to these torrential rains that caused rock surface failure across a joint, fracture, and cleavage planes. One of the incidents was a landslide due to massive rain and winds that struck Penang Hills, a world-famous heritage landmark on 5th October 2017. Train operations were disrupted as earth mounds sunk near the middle station on the trail, and many assets such as transport cables and two funicular trains were lost. Also, around 4th and 5th November 2017, the island experienced an extreme storm that caused numerous landslides that damaged lives and property in Penang Island (Pradhan and Lee 2010).

Table 1 Past landslide events
Full size table
Fig. 1
figure 1

Landslides activities of Penang Hills (a–c), the landslides, and downpour that caused more than 100 homes inundated in Penang (d) and The Tanjung Bungah landslide and Paya Terubong (e)

Full size image

This research is aimed at applying the technique of PS-InSAR to measure ground deformations and ER to characterise underground water distribution, spatial stratigraphy, and material properties. The objectives include detecting ground surface displacements, imaging subsurface structures, and investigating saturated zones, and finally validating the output using borehole data. The in situ measurement for this study was carried at Paya Terubong, Malaysia, in 2019.

Study Area

This study was performed in Penang Island, northwest of Peninsular Malaysia, with the geographical coordinate (05° 41′ 64″ N, 100° 33′ 22″ E) as presented in Figs. 1b and 2a. Penang is a developed state, and its growth and industrialisation draw people worldwide (Paradella et al. 2015). The area is vulnerable to severe incidents such as subsidence, landslides, and collapse of human-made buildings because of the hilly nature of the region coupled with the highest rainfall of 647 cm per year (Gin 2015). Road cuts for modern buildings and deforestation activities on hills coupled with natural erosion are the leading causes of soil instability in the area (Chee et al. 2017). Paya Terubong is used as the study site for in situ measurement, as depicted in Fig. 2c. The topography of the area based on the digital elevation model is shown in Fig. 2d. The elevation ranges from 700 to 1000 m above sea level (asl) with the highest in the northern part, and lowest in the western region, along the road. The instability in the topography correlates with the boundary of the two classes of granite that makes up the island.

Fig. 2
figure 2

The geographic location and topography of the study area in Penang Island, Malaysia

Full size image

Geology of Penang Island

The knowledge of the geology of the study area cannot be overlooked. It is used in the resistivity analysis for the investigation of the subsurface activities (Ahmad et al. 2006). Penang Island is relatively stable and underlain by an igneous type of rock when classified based on the fraction of compositions of alkali feldspar to total feldspars, as shown in Fig. 3 (Kwan et al. 1992; Wilkinson et al. 2014). It consists of two districts, namely Northeast and Southwest, as illustrated in Table 1, and the geological status of the two regions differs though they are in the same state. Forty per cent of the northeastern part of the island is covered by the intermediate microcline granite (Ahmad et al. 2006). This granite is divided into three units, namely Ferringhi granite, Tanjung Bungah granite, and Muka Head microgranite, as shown in Fig. 3. These units display crustal interactions split by shear zones, faults, and veins. The southern part consists of a gentle slope with an average height of 360 m asl, covered by microcline granite, which is divided into Batu Maung and Sungai Ara granites, as presented in the same Fig. 3. Eighty per cent of this southern part is covered by Batu Maung granites (e.g. megacrysts), while the remaining twenty per cent is covered by the Sungai Ara granites. The alluvial deposit comprises clay, sand, gravel, and silt clay up to 20 m deep, and dominates the western region of the island (Hassan et al. 2018).

Fig. 3
figure 3

Map of the geology of Penang Island

Full size image

Soil Properties

Understanding the characteristics of the soil sequence is essential to classify its impacts. The soil series of Penang Island is composed of residual soil as categorised under granitic soils with an average of dry density varying between 1.41 and 2.31 kkN/m3 and depth to bedrock ranging from 2.41 to 40 m (Ahmad et al. 2006). This soil series is formed by the deposit of organic material that is not transported. The soil series of Penang Island is classified into eight types, as follows Steep land, Urban land, Serong, Rengam, Local alluvium–colluvium association, Kuala-Kedah Permatang association, Selangor Kangkong association, and Rengam Bukit Demian association (Pradhan and Lee 2010). The steep soil type dominates the major part of the island, and this is where most of the landslide accidents have been reported (Ali et al. 2011). The distribution of the soil series of the study area is presented in Table 2. Penang Island consists of two types of granite rocks based on their differences in age, texture, and mineralogy, i.e. Type I: Sungai Ara (southern part), and Type II: Bukit Bendera (northern part) (Tan 1994; Hashim et al. 2018) as shown earlier in Fig. 3. Type II is found in the southwestern (Batu Maung) part of the island. The main fault that passes from the north–south divides the island into two vertically. This fault extends to other areas like Tanjung Bunga, Paya Terubong, Tanjung Bunga, Air Itam, etc., that is perhaps one of the causes of periodic landslide situations in these areas shown in Table 2.

Table 2 Series of main soils of Penang Island (Pradhan and Lee 2010)
Full size table

Materials and Methods

PS-InSAR technique has been applied for the detection of landslides and active landslide areas on the island of Penang in this work. The technique is a cost-efficient tool for remote sensing. It provides ground deformation measurements with high spatiotemporal resolution as compared to other conventional devices like Global Positioning System (GPS), Levelling, and Global Navigation Satellite Systems (GNSS) (Chen et al. 2017). The technique has been commonly used to measure the slow movement of landslides and applied to ground deformations measurement (Tong and Schmidt 2016; Chen et al. 2018). It is applied to process S-1A data in this study for ground deformation detection.

S-1A is one of the many image-capturing sensors operating in the C-band with a revisiting period of 12 days. It was launched on 3rd April 2014, by the European Space Agency (ESA) and operated in dual-polarisation (Navarro-Sanchez et al. 2014). It is a promising sensor in InSAR application due to its short revisit time (Haghighi and Motagh 2017). This mission offers four types of services, namely Interferometric Wide (IW) Swath, Strip-map (SM), Wave (WV), and Extra Wide Swath (EW) with different polarisation conditions (Devanthéry et al. 2016), as shown in Table 3.

Table 3 Mode of acquisition of Sentinel-1A data
Full size table

Similarly, ER method is a widely used method. It is an inexpensive, non-invasive, and fast method that generates appropriate information about subsurface settings. It is sensitive to the water content of the soil layers and good to examine potential landslide areas (Wu et al. 2018). The spatial distribution of resistivity in the subsurface is determined via the inversion model, which explains the electrical potential measured. There exist various types of methods, but the least square is commonly used. This approach offers a 2-D ground resistivity and allows identification of anomalies in the ground. Various applications of ER include studying weathering (Bièvre et al. 2016), investigating underground natural resources, and identification of the internal structure of various slopes deformations (Di Traglia et al. 2018).

As mentioned previously, this study used PS-InSAR, a ground motion detection tool, to track Penang Island landslides. The method relates slave scenes over the same region to one suitably chosen master. PSI-SAR is more flexible than traditional InSAR and had addressed the uncertainties that may be introduced to the results by atmospheric conditions on the radar pulses and ground reflectivity variation in ground vegetation or moisture content (Ferretti et al. 2000; Hung et al. 2011). The InSAR technique generally involves multiplication of two or more images of SAR to determine distances between the moving antenna and target area to get amplitude and phase information (Perrone et al. 2006). A radar system sends the signal to the surface of the earth where the antenna receives and records the backscattered signals from the target. The two ways pulses between the radar and the calculated target area are called an interferogram that contains information about the stability of the surface of the earth (Lingyun et al. 2013).

The multi-temporal (MT) analysis was performed over the study area by exploiting the available dataset composed of a single look complex (SLC). The acquired 60 and 56 S-1A images on (20th February 2017, and 13th November 2019), and (1st September 2017 and 30th December 2018), from ascending and descending tracks were processed using SARPROZ software to measure, monitor, and map ground deformation in the study area. The S-1A data were acquired using Terrain Observation by Progressive Scans (TOPS) imaging mode operated by the ESA (Razi et al. 2018). The satellite used for data acquisition works in IW mode by passing over the area of interest (AOI) north to south in descending pass direction and south to north in ascending pass direction with 12 days revisiting cycles. In IW Swath Mode, data are acquired through a series of bursts with mutual overlaps by regular switching antenna beam between the sub-swaths (Roccheggiani et al. 2019). Table 4 presents some features of the S-1A dataset.

Table 4 Descriptions of the dataset of Sentinel-1A used for this research
Full size table

The first step was to create a project directory named SLC, in which the images downloaded were copied and stored. All the images were imported into the software, and appropriate AOI, such as latitude, longitude, and radius, shown in Table 5, were selected (Maghsoudi et al. 2018). The images acquired on 16th April 2018, and 29th March 2018, were chosen as the master images for the ascending and descending tracks, respectively, based on many crucial components such as fair-weather conditions, doppler vector, and normal temporal base (Perissin and Wang 2012; Razi et al. 2018; Luo et al. 2017). The images were later co-registered with the master images without any spatial errors to ensure that each target has azimuth and range geometry for master and slave scenes (Bakon et al. 2015) as depicted in the star graph of Fig. 4. The reflectivity map and amplitude stability index (ASI), sparse point selection mask, ground control points (GCP) selection, and external DEM and synthetic amplitude in SAR coordinate parameters were chosen for PS Candidates (PSC) (Declercq et al. 2017). Such steps offered opportunities for creating a map of reflectivity and displaying the DEM in the AOI. Due to the presence of additional noises associated with the hilly and vegetated areas, both the Goldstein filtering and multi-look (ML) modes were used to filter the data (Bakon et al. 2015). The images were geocoded to Google Earth and projected to 3-D orthophoto maps for the exportation of the results. The chart showing the processing of S-1A imageries using SARPROZ software is presented in Fig. 5.

Table 5 Subset of the dataset
Full size table
Fig. 4
figure 4

Star graph network configuration of master and slave pair images of the Sentinel-1A time series. Each connection represents an interferogram (a phase difference map) considered to retrieve the SAR phase variation over time for each highly coherent ground target

Full size image
Fig. 5
figure 5

Flowchart of the SARPROZ software for processing S-1A data

Full size image

The resistivity survey was done when the current electrodes were inserted into the ground, and the potential difference was measured through the potential electrodes. A dipole–dipole configuration was used as it offers good lateral resolutions as compared to other arrays such as Schlumberger and Wenner arrangements (De Bari et al. 2011; Wu et al. 2018; Kannaujiya et al. 2019). The set-up (Fig. 6a) includes ABEM Terameter SAS 4000, electrodes, cell, and reels of cable in the field. Two resistivity profiles were laid for imaging subsurface using 41 electrode poles with a spacing of 2.5 m. The data acquired were processed with the application of surfer software via computer iterations, where vertical 2-D resistivity contour maps were generated along the direction of the electrodes. The result was validated using Ahmad et al. (2014) borehole measurements.

Fig. 6
figure 6

Electrical resistivity set-up

Full size image

Results and Discussion

The interferograms were generated from the 60 and 56 master–slave pairs of S-1A images acquired between 20th February 2017, and 13th November 2019, and from 1st September 2017 to 30th December 2018, for both the ascending and descending tracks, respectively, based on PS-InSAR technique. Due to the presence of additional noises associated with the hilly and vegetated areas, the multi-look factor and Goldstein filter were used to filter the data.

The quality of the generated interferograms depends on the temporal and normal baselines. It provides the ability to examine some smaller areas with significant problems of land deformation using ground deformation processing. A histogram graph shown in Fig. 7 is suitable for inspecting the isolated areas as it gives a better understanding of the statistical overview of connections coherence and information of PS points used for land displacements interpretations and analyses. The points values of 35,961 and 37,936 were obtained for the two tracks, respectively. These points were used to calculate the time series. The linear trend and height correction between neighbouring PS candidates (PSC) was estimated by applying a temporal coherence histogram in the time domain after the estimation of all the parameters and APS removal. The number of PS points decreases with threshold values increasing. As illustrated in Figs. 3 and 6, Paya Terubong is situated in the centre of the island where extensive landslide tragedies had occurred.

Fig. 7
figure 7

The temporal coherence histogram of points after parameter estimation and APS removal

Full size image

Interest in unstable terrain identification increases as many neighbouring points display coherent and temporarily progressive motion. The ground displacement of the PS distribution points in the RGB colour scale is presented in Fig. 8a, where red points depict extreme movements with the highest displacement and correspond to unstable targets showing vertical motions inferior to ± 12 mm/yr from the satellite LOS. The red dot band coincides with the areas of deformation where the instability is readily observable. Separate-colour points correlate with different PS displacement values and, many of these points are on an unstable slope amid the mountainous terrain. The figure generally shows that Penang Island area is globally unstable, where the blue colour connotes moderate displacement points throughout the study area and is used to discriminate movements beyond the observable background noise. The green colours denote stable points, whereas the densely distributed yellow points refer to the ground’s vertical motion, representing unstable SAR backscattering targets positioned on stable terrain.

Fig. 8
figure 8

Geocoded PS points on Google Earth and Time series of land deformation starting from 14th February 2017 to 1st September 2018 represented by PS point (ID 19641)

Full size image

These results of PSI processing were geocoded and overlaid with the 3D orthophoto land mapping for validation and ground visualisation. The number of PS points increases as the threshold value decreases. From the InSAR processing, the coherence values of the PS points range from 0.2 to 0.65. The low value of this parameter is due to the presence of vegetation covers. The time series displacement for a PS point of Paya Terubong site is presented in Fig. 8b, where the x and y axes represent the temporal baseline in days and deformation value in millimetres, respectively. The movements exhibit linear motion ranging from − 1.5 to 41.3 90 mm/year away from a LOS of the system. The steep slope shows that there is a high acceleration of land movement. Different colours of the image correspond to a different displacement value of PS.

The results of the geoelectrical survey conducted to investigate the spatially varying ground movement about the geological setting of the affected area, as detected by the InSAR technique, are presented in Fig. 9. A significant contrast was not noticed from the results implying there are no geological discontinuities of the study area. The results only indicated some locations characterised by low resistivity values, which could be associated with the presence of water. In contrast, the high rate of resistivity may be related to the existence of underground events.

Fig. 9
figure 9

The electrical resistivity results obtained with the dipole–dipole method (a represents line 1 and b depicts line 2)

Full size image

These results reflect the presence of underground water that typically alters the alluvial deposit, as shown in Fig. 9a and b. The observed low resistivity zone of 0–20 Ωm at a depth of up to 5 m at the study area coincides with the value of bedrock, according to the reports of (Ahmad et al. 2014). High resistivity values were observed at a depth of 5 m and above, indicating the presence of structurally shaped boulders (bedrock units) due to weathering and erosion (Ahmad et al. 2006). These results suggest that the bedrock units are situated at a depth of around > 5 m depth, the resistive layer is higher at greater depths. The gravelly terrace deposits of surveys (1 and 2) in Fig. 9 are imaged with two resembling layers having resistivity values between 3000 and 5000 Ωm. The data obtained using the dipole–dipole array, as seen in Fig. 6, are noted to be responsive to variations in lateral resistivity. Such variations are regarded as the ground deformation observed using the InSAR technique.

Discussion

It is evident from the overall results of PS-InSAR and ER that different factors influence the deforming areas investigated. The deformation pattern identified in Paya Terubong is regarded as being due to the reactivation of past landslide phenomena. The tropical depression predicted by the South China Joint Typhoon Warning Center (JWTC) triggered a heavy rainfall that reactivated landslides located in a large region of Penang and Kedah and led to major disasters in those places (Meteorological Malaysia Department 2017). The observed ground deformation may be attributed to a landslide occurrence for which, however, the only indication noticeable in the field and ER results could be related to the presence of very low resistivity values detected shown in Fig. 9. The knowledge of deformation mechanisms is summarised as probably attributable to human-made settlements on the ground. There can be no conclusive results on the origin of the deformation; however, the details presented by both PS-InSAR and ER techniques helps to predict a slow-motion subsidence phenomenon. The detected displacement could include, the shallow layer of unconsolidated sediment, which leans the conductive clay material. We cannot rule out the common triggering factors observed in the overall deformation processes. The findings of this study preclude us from having a definitive interpretation. The triggering factors on ground motions could include the shift in the geomorphological setting, subduction, sandy soils, and changes in groundwater levels besides those they have been previously mentioned.

Validation of Electrical Resistivity Results using Borehole Analysis

There is a need to validate the results of electrical resistivity while giving a detailed analysis. This validation was achieved with the application of borehole analysis of the study area.

Borehole logs collected from the Department of Public Works, Penang, in the work of Ahmad et al. (2014) were analysed and summarised to verify and validate the resistivity findings. The interpretations were based on the descriptive statistics of the thickness of sand and silt. Table 6 summarises the number of boreholes obtained, number of locations, and mean values of the thickness of both the sand and silt of the areas. In five areas, 112 borehole data were acquired from twelve different locations. The maximum sand thickness measured was attributed to its proximity to the sea, together with the area’s reclaiming activities. In Paya Terubong, the lowest silt thickness is 4.45 m. This value refers to the saturated areas, measured using the electrical resistivity method.

Table 6 Summary of the number of boreholes obtained, number of locations and mean values of the thickness of both the sand and silt in the study areas
Full size table

Conclusion

The past and potential ground deformation events in Penang Island have been monitored and mapped in this research with the application of the PS InSAR technique and electrical resistivity measurements.

The InSAR results suggested that the rate of landslides’ magnitude is constant with a velocity of mm/yr. The areas with ground deformations (subsidence and uplift) were revealed through the InSAR technique, while the geophysical method revealed likely triggering mechanism in the subsurface. The results of this work are in line with previous studies to show that the middle of the island is vulnerable to landslides owing to its hilly terrain and dense forests. Paya Terubong remains the region that has had the most significant number of recent landslide occurrences in Penang Island.

The resistivity data obtained at a construction site located along the hillside in Paya Terubong were made possible, because the construction site had already deforested, flattened, and cleared the site for accessibility. Good results about the subsurface region were achieved from the 2-D resistivity data in which certain saturated locations were identified, examined, and interpreted. The aquifer was considered for areas with low resistivity values. Landslide events are also suspected as a result of certain boulder irregularities observed in the subsurface that may have developed as a result of weathering processes. These granitic boulders could be triggered by vibrations of pieces of machinery at the construction site even though mixtures are made a few metres away from the construction site.

Since there is no monitoring system in existence to study the dynamics of landslides in Penang and Malaysia as a whole, the integration of these two techniques is useful for landslide mapping and monitoring as it has comprehensive coverage with a high spatial resolution at a low cost. One single approach cannot be used as a stand-alone tool for solving landslide challenges in an urban area. The combination of the two techniques thus allowed the detection of certain factors that could cause landslides in the region. This work can be used as a guide for detecting possible areas of landslides. The generated map will assist the government and contractors make suitable decisions in the expansion of Penang Island.