Introduction

Nature’s fury is frequently presented before us in the form of devastating earthquakes. A major earthquake struck Nepal on 25 April 2015 at 11:56 (local time), (28.147°N; 84.708°E) of Mw 7.8 (focal depth ≈ 8 km), (at 34 km ESE of Lamjung, Gorkha district and approximately 80 km northwest of Kathmandu). Later, on 12 May 2015 at 12:50 (local time) another major earthquake (27.837°N; 86.077°E) of Mw 7.3 (focal depth ≈ 15 km) struck Nepal near Dolakha (NE of Kathmandu and SE of Kodari). The epicenter was on the border of Dolakha and Sindhupalchowk districts. So far more than 316 numbers of aftershocks (of magnitude ≥ 4) until 14 June 2015 (Source: NSC) have already occurred (Fig. 1). It may be observed that more number of aftershocks are located towards eastern side of the Kathmandu. The Nepal earthquake has caused about 8622 loss of human lives and more than 16,808 people were injured. There are more than 71 deaths in India too. These two major earthquakes have also caused wide-spread landslides covering large area. Most of the landslides have occurred on higher Himalayan ranges and the landslide zone follows the regional WNW–ESE trending Himalayan structural features (Fig. 1). This study pertains to detection of earthquake induced landslides and existing landslides through satellite images and field observations.

Fig. 1
figure 1

Major tectonic features of the central Nepal Himalaya, location of major earthquake epicentres (filled stars) and aftershock (filled circles) up to 14 June 2015

Full size image

The focal mechanism of the main shock, the 7.8 and 7.3 magnitude earthquake suggests the rupture occurred on a shallow dipping (5°–10° to North) reverse fault aligned with the main range front (USGS 2015). This mechanism, together with the depth of the hypocenter at about 8-15 km, indicates that the event ruptured the Himalayan frontal decollement (Main Himalayan Thrust), the major fault that accommodates deformation at the front of the Himalaya. It is also reported that the rupture did not reach to the surface.

The rupture of this earthquake initiated in an area about 70 km northwest of Kathmandu, with slip on a northward shallow dipping fault and rupturing the rocks eastward by some 200 km and southward by around 110 km, breaking a fault segment some 22,000 sq km in area, with as much as 4 m slip in places. Due to this earthquake huge volume of Himalayan rocks suddenly got accelerated and staggered over the Indian Shield for about 4 m slipping over the earthquake rupture plane. InSAR studies also suggested a maximum displacement (> 1.2 m) moving toward the satellite is observed around 30 km east from Kathmandu (Geospatial Information Authority of Japan - http://www.gsi.go.jp/cais/ topic150429-index-e.html). Peak ground acceleration (PGA) generated during the 7.8 Mw main shock of each horizontal component was about 0.16 g and the vertical PGA was around 0.19 g (www.strongmotioncenter.org).

Earthquakes, generally of magnitude 6.0 and above, induce ground intensities ≥ VI, trigger landslides in the critical reaches of hill slopes. In this study, all available pre- and post-earthquakes satellite datasets from Landsat-8 (USGS), high resolution images from Google Earth have been utilized to identify earthquake induced landslides. Intermittent cloud covers have repellently hampered satellite image interpretation work related with Nepal earthquakes.

A sum total of 15,551 numbers of landslides have been identified, out of which 3429, 5463, 1700, 1479 reported to have occurred in the Gorkha, Sindhupalchowk, Rasuwa and Dolakha districs of Nepal respectively (Martha et al. 2017). In this study Landsat-8 pre- and post-earthquake data used for identification of earthquake induced and enhanced landslides; total six major landslides areas have been identified in the post-earthquake Landsat-8 image (Fig. 2). Further, pseudo colour transform (PCT) technique which is developed by Saraf (1998a, b), applied for earthquake damage assessment has successfully been employed on Landsat-8 pre- and post-earthquake datasets to accurately identify the landslides. Many such landslides have also been verified during field investigations.

Fig. 2
figure 2

Landsat-8 post-earthquake image of 1 June 2015 illustrating the location of various landslides induced by the Nepal earthquake of 25 April 2015 and 12 May 2015. The yellow colored boxes show the area maximum affected by landslides and details are shown in the following figures (color figure online)

Full size image

Data Used and Methodology

Landsat-8 datasets were obtained from the USGS (http://earthexplorer.usgs.gov). It is an American Earth observation satellite and carries two instruments: Operational Land Imager (OLI) and Thermal Infrared Sensor (TIRS), it consists of nine spectral bands with a spatial resolution of 30 m for band 1–7 and 9. The resolution of band 8 (panchromatic) is 15 m. Thermal bands 10 and 11 have a resolution 100 m.

Among the available Landsat-8 pre- and post-earthquake images of two different dates (11 April 2014 and 1 June 2015) have been used to study the Nepal earthquake induced terrain changes.

Several image change detection methodologies are available which can be categorized as pixel-based wherein an image pixel is the fundamental unit of analysis and the second one is the object-based method, in which image objects are first created prior to further analysis. Different techniques for image based change detection studies are as discussed below.

Pixel-based change detection techniques use spectral characteristics of an image pixel to detect and measure changes without considering the spatial context. Detailed on pixel-based methods are explained by various researchers (Coppin et al. 2004; Deer 1995; Ilsever and Ünsalan 2012; Lu et al. 2004; Singh 1989; Martha et al. 2012).Different pixel based change detection techniques are enumerated below.

Image differencing is executed by subtracting the pixel value (a digital number) of a given band belonging to one date from the pixel value of the same location for the same band of another date resulting into a new image. Changed feature (pixel based) from unchanged features are determined by a threshold value based on standard deviations. Under the image ratioing technique ratio between two co-registered images is taken to obtain the change results and this method can handle calibration (including sun angle, shadow and topography impact) errors efficiently (Rignot and van Zyl 1993). Image regression is performed by assuming one image (master) is assumed to be a linear function of the other image (reference). Then the spectral signature of master image is adjusted to match with that of the reference image by using least-squares regression analysis (Lunetta et al. 2004). Normalized Difference Vegetation Index (NDVI) is represented by a ratio between the spectral band difference and addition as NDVI = (NIR − RED)/(NIR + RED) where NIR is the near-infrared band and RED is the red band response. Once NDVI is computed using images of two dates then differencing or ratioing are applied to determine the change (Nelson 1983; Martha et al. 2010). Under change vector analysis pixel values are taken as the vectors of spectral values of multiple image bands and change vector is computed by subtracting vectors for all pixels at different dates (Malila 1980). Change is determined from the direction of new vector while the length corresponds to the magnitude.

Transformation category includes principal component analysis (PCA), Kauth–Thomas transformation (KT)/tasseled cap transformation (TC), Gramm–Schmidt (GS) and Chi square transformations. These methods have capability in reducing data redundancy between bands by transforming multivariate data to a new set of components (Lillesand et al. 2008). PCA assumes that areas of change are not highly correlated and the covariance matrix is used to transfer data to an uncorrelated set. The first principal component contains most of the data variation. The next largest amount of variation is governed by the successive principal component and is orthogonal to the preceded one. Change detection is applied after acquiring the principal components. The KT/TC method involves fixed linear transformation of a multi-date and multiband dataset. The spectral data is the function of a particular characteristic of scene classes. Change is indicated by the brightness, greenness and wetness values (Lu et al. 2004).

Object-based change detection wherein objects are segmented by thresholding of multi-temporal images and change is analyzed based on average spectral band values of the object or by extracting features (e.g. geometry and texture) from the original object. In image-object change detection technique changes are detected by comparing the following properties such as geometrical (width, area and compactness), or spectral (mean band values), and/or extracted features (e.g. texture) of the image objects (Lefebvre et al. 2008; Hall and Hay 2003). Under the class-object change detection approach extracted object is categorized into a specific class and changes are detected by comparing the independently classified objects (based on both texture and spectral information) from multi-temporal images. Multitemporal-object change detection technique involves generation of spatially corresponding change objects as image-objects from different dates through segmentation. The algorithm for this purpose does segmentation of multi-date image-set and computing its spectral features (i.e. standard deviation and mean) from each date for all image objects (Desclée et al. 2006). Further, Singh (1989) has provided detailed review on various image change detection techniques.

In the present study PCT change detection technique has been employed because various worker e.g. Sarkar and Saraf (1999) and Saraf (2000) have found this technique very simple and highly effective in case of detecting earthquake induced landslides.

False colour image of Landsat-8 datasets was generated to achieve higher contrast level among the different features. The PCT technique applied to enhance the ground damages and assess earthquake-induced landslides. The conventional PCT technique facilitates the combination of various bands in different colour schemes, such as red, green and blue (RGB), intensity, hue and saturation (IHS) and cyan, magenta and yellow (CMY). So to apply the PCT technique the post-earthquake scene is kept in the red channel and the pre-earthquake scene in the green and blue channels. This has provided a change index with red depicting positive change in terms of reflectance between two dates, blue and green depicting negative change, and black and white depicting no change (Sarkar and Saraf 1999; Saraf 2000; (Figs. 3, 4, 5). The Pseudo Colour Transformation (PCT) technique used to clearly identify the earthquake induced landslides and reactivated landslide by earthquake (Saraf et al. 2009). In this work nearest available pre and post-earthquake satellite images have also been extracted from Google Earth to draw correlation among the landslides of different time frame.

Fig. 3
figure 3

(I): Satellite image of Langtang valley (box 1 of Fig. 2). a Pre-earthquake Landsat-8 image of 11 April 2014 exhibiting a few small brownish areas depicting existing landslides; b Post-earthquake Landsat-8 image of 1 June 2015 illustrating major new landslides triggered by the Nepal earthquake of 25 April 2015 and 12 May 2015 (shown by an arrow); c PCT image of the Langtang valley (near Langtang Khola) as shown in (a) and (b). Red areas depict earthquake induced landslides and other ground changes; d Masked image of (c) showing major landslides. (II): Rasuwa area (box 2 in Fig. 2) a Pre-earthquake Landsat-8 image of 11 April 2014. There are a few small brownish areas depicting existing landslides; b Post-earthquake Landsat image of 1 June 2015 illustrating new landslides triggered by the Nepal earthquake of 25 April 2015 and 12 May 2015; c PCT image of the Rasuwa (Near Pasang-Lhamu way) shown in (a) and (b). Red areas depict earthquake induced landslides and other ground changes; d Masked image of (c) showing major landslides (color figure online)

Full size image
Fig. 4
figure 4

(I): Dandagaun area (box 3 in Fig. 2) a Pre-earthquake Landsat image of 11 April 2014. There are a few small brownish areas depicting existing landslides; b Post-earthquake Landsat image of 1 June 2015 illustrating major new landslides triggered by the Nepal earthquake of 25 April 2015 and 12 May 2015; c PCT image of the Dandagaun area shown in (a) and (b). Red areas depict earthquake induced landslides and other ground changes; d Masked image of (c) showing major landslides. (II): Kashigaun area (box 4 of Fig. 2) a Pre-earthquake Landsat image of 11 April 2014; There are a few small bluish-white areas depicting existing landslides; b Post-earthquake Landsat image of 1 June 2015 illustrating major new landslides triggered by the Nepal earthquake of 25 April 2015 and 12 May 2015; c PCT image of the Kashigaun (near Buri Gandaki River) area shown in (a) and (b). Red areas depict earthquake induced landslides and other ground changes; d Masked image of (c) showing major landslides (color figure online)

Full size image
Fig. 5
figure 5

(I): Gorkha area (box 5 of Fig. 2) a Pre-earthquake Landsat image of 11 April 2014. There are a few small brown areas depicting existing landslides; b Post-earthquake Landsat image of 1 June 2015 illustrating major new landslides triggered by the Nepal earthquake of 25 April 2015 and 12 May 2015; c PCT image of the Gorkha area shown in (a) and (b). Red areas depict earthquake induced landslides and other ground changes; d Masked image of (c) showing major landslides. (II): Bahrabise landslide area (box 6 of Fig. 2) a Pre-earthquake Landsat image of 11 April 2014. There are a few small brown areas depicting existing landslides; b Post-earthquake Landsat image of 1 June 2015 illustrating major new landslides triggered by the Nepal earthquake of 25 April 2015 and 12 May 2015; c PCT image of the Bahrabise landslide area shown in (a) and (b). Red areas depict earthquake induced landslides and other ground changes; d Masked image of (c) showing major landslides (color figure online)

Full size image

Satellite Data Based Landslide Studies

Satellite images covering different parts of landslide affected areas of central Nepal region revealed several landslides of various dimensions. Landslides have been identified on satellite images as shown in Fig. 2 (area bounded by boxes) and analysis has been carried out to show only landslide affected areas. Out of many landslide zones some zones as marked as boxes 1–6 as shown in the Fig. 2 have been selected based on availability of cloud free and good quality pre and post earthquake satellite images. These areas are 1-Langtang valley (Fig. 3-I), 2-Rasuwa (Fig. 3-II), 3-Dandagaun (Fig. 4-I), 4-Kashigaun (Fig. 4-II), 5- Gorkha (Fig. 5-I), 6-Bahrabise (Fig. 5-II).

Field Based Landslide Studies

During field investigations the following three different sectors have been covered and shown in Fig. 6, these are (a) towards east of Kathmandu; (b) towards northwest of Kathmandu; (c) towards southwest of Kathmandu. It is also observed that the tectonic features such as faults have also controlled occurrence of landslides. This effect has especially been observed in two zones and discussed below. Figure 7 shows tectonic features of central Nepal region overlaid on satellite image along with the locations of most landslide affected regions.

Fig. 6
figure 6

Field work locations (shown as landslide locations) in and around Kathmandu, Nepal

Full size image
Fig. 7
figure 7

Satellite image of the central Nepal region showing locations of prominent landslides and tectonic features. MCT Main Central Thrust, MBT Main Boundary Thrust, MFT Main Frontal Thrust, EL Everest Lineament, GL Gaurishankar Lineament. Faults and Tectonic features are taken from GSI (2000)

Full size image

Eastern Sector

This sector covers Sindhupalchowk and road section to Bahrabise (Fig. 6). In this sector several landslides could be observed including road subsidence on Araniko highway between Dolalghat-Kalika section which runs almost parallel to Sun Kosi river. These were induced due to heavy rain fall and might have been triggered by 12th May 2015 earthquake as this earthquake was located nearer. Huge landslide has been observed on Kathmandu-Bahrabise road at latitude 27.76° and longitude 85.87° which has brought lot of debris comprising very large to large rock pieces and small pieces of crushed rocks down the slope. Geologically this slope is comprised of layered rocks (predominantly phyllites) dipping into the hill. Several other landslides have also been observed on satellite images in the area north of the above landslide i.e. Ramche-Bahrabise.

This area is bounded by MCT on northeast, northwest, southwest and to southeast by the Gaurishankar Lineament (Fig. 7). Although no significant tectonic feature occur along the rivers course at the landslide site and appear to be controlled by a fracture line. Moreover, the thrusting effect of MCT might have weakened the hill ridges mostly along the river valley (Fig. 7).

On Araniko Highway several medium sized fresh landslides have also been observed (Fig. 8-I, II).

Fig. 8
figure 8

(I): Pre- and post-earthquake satellite images depicting origin and growth of earthquake induced landslide (debris/soil fall) near Zero Kilo village on Araniko Highway along with field photographs. (II): 12th May 2015 earthquake induced landslide (near Dolalghat village) on Araniko Highway

Full size image

Northwest Sector

This sector is comprised of Bidur and road sections to Dhunche following the Trishuli river course (Fig. 9-I). Out of several landslides a large scale landslide has been encountered along the Trishuli river southwest of Dhunche.

Fig. 9
figure 9

(I): Landslide on the Dhunche road (before Ramche). Existing large landslide of about 1 km long (between Laharepauwa–Ramache villages) further reactivated by Nepal earthquakes consisting large blocks of schist, gneisses and soils (taken from Google Earth and Google map). (II): Field photographs of existing large landslide of about 1 km long (between Laharepauwa–Ramache villages) further reactivated by Nepal earthquakes consisting large blocks of schist, gneisses and soils

Full size image

This landslide is also extensive and affected the N-S oriented concave inward mountain slope. This ridge essentially is made up of schist and gneiss rocks and slide initiated from the height of 2450 m and extends up to the Trishuli river at an altitude of 750 m damaging the road at an altitude 2020 m. During the investigation it has been observed that the landslide is perennial in nature as old slided slopes still exist. However the slide zone got reactivated during the earthquake (Fig. 9-I, II).

Further, a NNE–SSW trending fault has been identified as per the atlas published by GSI (2000) occurring in this region (this fault can be seen in Fig. 7 located just west of Dhunche). This fault controls the Trishuli river course and effect of faulting appeared to be manifested intensely in the slide affected hill slope as indicated by high degree of shattered rocks.

Southwest Sector

This sector is along the road from Kathmandu to Hetauda at the foot hills of Nepal Himalaya. Quite a few landslides have been observed and among them at least one is of larger size. However, on the other side of the river valley hill slopes have been affected by several landslides (Fig. 10-I). Incidentally NNE-SSW trending fault passes through this region along the valley controlling the course of river (Fig. 10-II, this fault can be seen in Fig. 7 located north of Hetauda). This fault has weakened the rocks of the hills and evidence of faulting could be observed at one location as shown below. Figure 11 show extensive landslides zone along the Bagmati river valley and on Kathmandu-Hetauda road near Shital-chowk village of Chhaimale VDC.

Fig. 10
figure 10

(I): Landslides zone along Bagmati river valley and on Kathmandu-Hetauda road near Shital-chowk village of Chhaimale VDC. (II): Faulted river valley. Two small ridges as shown by the arrow got detached and subsided in the river valley due to faulting

Full size image
Fig. 11
figure 11

Large landslide occurrence along Bagmati river valley and on Kathmandu-Hetauda road near Shital-chowk village of Chhaimale

Full size image

Conclusions

Large numbers of landslides have been identified including landslides triggered by the 25 April 2015 Gorkha earthquake and the 12 May 2015 Dolakha earthquakes, as well as reactivations of landslides that were present before the recent earthquakes sequence began. Approximately 1975 landslides identified and mapped using Landsat-8 satellite image Fig. 12.

Fig. 12
figure 12

This map shows approximately 1975 landslides which have been identified and mapped (shown as blue and green colors) using Landsat-8 satellite images and of 25 April 2015, 12 May 2015 Nepal earthquake (color figure online)

Full size image

In this study, Landsat-8 post earthquake image of 1 June 2015 shows the location of various landslides induced by the Nepal earthquake (Fig. 2). Major landslides are limited to a zone that runs east–west, approximately parallel to the transition between the Lesser and Higher Himalaya (following almost Main Central Thrust). Further it is important to note that most of the earthquake induced landslides have south/sun facing slopes. Time-series satellite images indicate that there were already initial signs of slope failure. Later this landslide became a major one as can be seen on the 10 August 2014 image, which is rain induced rather than the earthquake. Also, satellite image of pre- and post- earthquake (Fig. 8-I) reveals that there were already signs of slope failure before the 25 April 2015 and 12 May 2015 Nepal earthquake.

In general, locations that were affected by landslides after the Gorkha earthquake also re-suffered from widespread land sliding due to the Dolakha earthquake. These areas have suffered damaging levels of shaking in two successive large earthquakes, and must be considered as having a very high risk of failure during the 2015 monsoon season.

The present study successfully demonstrated that remote sensing techniques can provide a useful tool to map and delineate earthquake induced landslides and other damage to the ground. The technique developed to assess earthquake induced damage using high resolution remote sensing data is cost effective, unbiased, and free from subjectivity, time saving and provides quantitative damage assessment. Using the presented approach, the maximum relief and planning can be provided in the minimum time and in less accessible areas of the Himalayas.