Identification of Potential Natural Slope Failure Zones by Geomorphological Analyses Using Raster Slope Shading of LiDAR; Case Study from Kegalle, Sri Lanka

Abstract

There are three approaches to defining the potential instability zonation of natural slopes. The first approach is to understand the failure mechanism through soil properties in a slope, and the mechanisms of failure and movement. The second approach aims to understand the tendency of slopes to fail, with terrain factors that form the essential characteristics of slopes. Understanding the spatial distribution of slope failures and their patterns on a particular slope is the third approach and the focus of this research. All approaches require field verification with expert knowledge.

Slope failure zones contain unique topographic patterns that can be used to identify the failure shape and its dimensions. Slope failures are one of many natural denudation processes. Most slope failure landform units, considered as past slope failures tend to expand naturally. If there is no human involvement, zones of past natural slope failure can be also categorized as potential zones of future slope failure. The large-scale geomorphological analysis is the best approach for clearly identifying landform units associated with potential zones of slope failures. The best scale is 1:10,000. Two-dimensional or three-dimensional raster interpretation of slopes can be used to visualize more clearly the actual shape of slope failures. For the study, raster geomorphological mapping uses LiDAR survey data to characterize the landform units of slope failures and to prepare a landslide susceptibility evaluation.

1 Introduction

Sri Lanka is an island near the equator in the Indian Ocean, with central highland terrains frequently subjected to slope failures due to highly intense rainfalls, especially during two monsoon periods. The failure-prone area has reportedly increased from 20% (13,000 km²) to 30% (19,500 km²) of the country’s total land area over the last two decades (LHMP Annual Report 2017). Every year, human losses and economic damage are recorded due to the expansion of human settlement and other activities into landslide-prone areas without land-use management practices (e.g., plantation, improper soil excavation). Some failures are reported as reactivations of past failures within the same terrain in the database of the National Building Research Organization.

All of these failures start with the reduction of the shear strength of an unstable soil mass. Shear failure along a slip surface occurs as shear strength values (resisting force) of slope materials become smaller than the shear stresses (driving force) acting on soil mass (Igwe 2014). Potential susceptibility and damage zones are defined based on understanding the failure mechanism of soil overburden and the downslope spreading mechanism (propagation). This process is one of the most widely used approaches to identify zones susceptible to failure initiation and for use in numerical downslope movement simulations.

Landslides are triggered by terrain factors that affect the shear strength of the soil and shear stress acting on the soil. Expressed as a function of failure initiation, these essential characteristics of slopes are geological, geomorphological, and associated with human activities (NBRO User Manual 1995). Another approach studies the process of failure initiation with terrain factors and is used for landslide susceptibility evaluation. Triggering events are various external factors (external stimuli) that change within short periods, such as rainfall and earthquakes.

Past failures in a landscape create unique topographic patterns and shapes that can be used to identify potentially unstable terrain. Geomorphological analysis of areas where particular landforms are identified leads to mapping past failures and failure zones. Most landform units derived from past slope failures tend to expand naturally. Consequently, areas of slope failures and colluviums identified by geomorphological analysis can be mapped as zones of potential slope failure. Through this approach, an understanding can be gained of the spatial distribution of failures and their patterns on a particular slope: used for evaluating future potential landslide susceptibility.

To best characterize and understand the typical landslide units, geomorphological analyses require high-resolution terrain data. Light Detection and Ranging (LiDAR) data with at least 2 m resolution is recommended for analysis of surface morphology (Trevisani et al. 2012; McKean and Roering 2004).

2 Study Area

Our study focuses on a 336,500 m² area with debris flow propagation starting as a slope failure recorded in Ganthuna Udagama, Kegalle in Sri Lanka. The study area coordinates are N7.130°, E80.430° to N7.136°, E80.435° WGS84 coordinate system with slope failure initiation location at N7.135°, E80.434° WGS84 coordinate system, at an elevation of 780 m above mean sea level.

Properties in the local community which sprawls downstream on gentle slope zones and a minor road were damaged by the failure (Fig. 1).

Fig. 1

Study area and the recent failure

Granitic gneiss and biotite gneiss with interlayered quartzite are the predominant metamorphic lithologies of the area. In the failure initiation area, the general foliation and lithological layers of granitic gneiss dip 40° with a dip direction of 245° (40°/245°). Two joint sets at 90°/050°/3 m⁻¹ and 60°/355°/2 m⁻¹ (dip/dip direction/intensity) can be observed (Source: LHMP, NBRO). Since the slope failure starts in the azimuth of 245°, the failure is geo-structurally controlled. Also, the debris propagation follows the general valley pattern pre-conditioned by geological structures shaped by natural denudation and erosion. With valley patterns and denudation processes mainly controlled by geological structures and weathering conditions of lithologies in this metamorphic terrain, geomorphological mapping of typical landform units and spreading patterns is considered the best approach to evaluate slopes for susceptibility to failure.

Compared to the general density of town areas in Sri Lanka the community density in the study area is low. However, settlements are confined to the valley bottoms and spread along the main highways and secondary roads (Fig. 2). Road patterns follow the surface morphology due to hilly terrain (Fig. 1). The study area mainly comprises tea-cultivated lands, pine-cultivated lands, degraded forest lands, bedrock outcrops, and human ettlements.

Fig. 2

Focused study area for the research (Google image): (a) before the failure 3/2016; (b) after the failure 2/2017

3 Methodologies

Traditional geomorphological mapping is a two-dimensional record of three-dimensional landform units recorded in the field. However, two-dimensional or three-dimensional raster interpretation of slopes can be used to geospatially visualize the actual morphological conditions more accurately without going to the field to survey.

In this study, high-resolution LiDAR point cloud data, processed with vegetation filtering facilitated the accurate identification of the morphological state of the landscape. The airborne laser scanned the ground surface at 100,000 to 200,000 samples per second (laser pulses were reflected from objects on the ground and travel time was converted to elevation) to obtain a geospatial dataset of all laser pulse returns from the ground (i.e., point clouds). Aircraft position was precisely measured so that the position and orientation of the laser were always known (JICA 2016). GPS base stations broadcast corrections to airborne GPS units (JICA 2016). The point clouds were generated from a Digital Elevation Model (DEM) of the bare terrain (surface morphology) after filtering the laser pulses reflected from vegetation, buildings, and infrastructure. A prepared DEM was then used to analyze the distribution of geomorphological features on the ground.

For this study, the slope-shading method in ArcGIS was used. Hill-shade analysis could also be used for this purpose, but this approach generated an azimuthal-biased morphological visualization and so was not used for this research. Preparation of the slope map from the DEM and shading was completed using the raster stretch and raster resample method. This analysis removed the azimuthal bias of hill shades from the slope map with artificial sunlight directly overhead with no azimuth. The change in slope gradient was depicted continuously from white to black by flat surfaces in white; low gradients in light grey; moderate gradients in dark grey; and very steep surfaces in black. A color-ramped, partly transparent DEM was overlayed on the slope shading map in ArcGIS to better visualize landform units with elevation differences. The slope map is generally the most important tool for interpreting landform units with slope failure initiations and deposits.

To capture the landform units of slope failures, both the manual method (with field and geomorphological identification experience), and an analysis of raster plan curvatures by slope map method can employed. However, the raster curvature analysis required expert knowledge to filter unnecessary zones not susceptible to failure (through field verification and filtering). For this study, the analysis of raster plan curvatures by slope map was used.

Raster modeling of slope morphology was represented as profile curvatures and plan curvatures (Peckham 2011; Pennock et al. 1987). The profile curvature represented the direction of the maximum slope. A negative value of profile curvature represented an upwardly convex slope; a positive value of profile curvature represented an upwardly concave slope, and the zero value represented a planar slope (Fig. 3). The plan curvatures represented a measurement perpendicular to the maximum slope (Rana 2006).

Fig. 3

Profile curvature landform representation

The negative values represented the sideward concave at the pixel; while the positive values represented the sideward convex at the pixel, and the zero value represented the straight slope (Fig. 4).

Fig. 4

Plan curvature landform representation

Landform units that were susceptible to failure could be detected based on both spatial plan curvature value distributions and the minimum slope gradient that was vulnerable to failure initiations. Zones susceptible to failures were determined using Eq. (1).

$$ {\displaystyle \begin{array}{l}\left(\mathrm{Plan}\ \mathrm{curvature}<0\right)\times \\ {}\kern1.5em \left(\mathrm{slope}>=\mathrm{minimum}\ \mathrm{slope}\ \mathrm{angle}\ \mathrm{for}\ \mathrm{failure}\ \mathrm{initiation}\right)\end{array}} $$

(1)

Landslide initiation required negative plan curvature values with the slope angle greater than the minimum slope gradient for the failure of the terrain.

4 Results

The 2 m × 2 m pixel resolution DEM (Fig. 5) of the study area is prepared from the LiDAR data (survey period: 12/2015-4/2016) collected by the Survey Department of Sri Lanka (JICA 2016) prior to the recent failure. The slope map of the study area is prepared from the DEM using the ArcGIS tools with slope shading and visualization stretched using a standard deviation of n = 3. A resample during display is undertaken using the bilinear interpolation for continuous data. The resulting visualization appears as continuous shading of slopes, that enables interpretation of landform units (Fig. 6).

Fig. 5

Digital elevation model of the study area (resolution of 2 m), prepared in ArcGIS

Fig. 6

Slope shading map of the study area (resolution of 2 m), prepared in ArcGIS (Invert color ramp)

A partly transparent DEM with a rainbow colour ramp is overlayed on the slope shading map to better visualize elevation differences of the terrain in the study area (Fig. 7). Past slope failures are depicted on the slope shading map overlain by the colour ramped DEM (Fig. 8). Six past failures can be identified using this method (Fig. 8). The recent failure is mapped as a reactivation of past slope failures (Fig. 9).

Fig. 7

Slope shading map overlayed by DEM (resolution of 2 m), prepared in ArcGIS

Fig. 8

Past failures captured from the slope shading map overlayed by DEM, prepared in ArcGIS

Fig. 9

Failure in May 2016 can be identified as a reactivation of the past failure

The plan curvature distribution of the study area prepared from the ArcGIS tools represents the propagation potential for convergence and divergence through the surface of the landform. The negative values of plan curvature (plan curvature <0) represent the sideward concave landforms and are expressed as topographic units of failure propagation. In the Ganthuna Udagama zone within the Kegalle District, a slope of 27° is the minimum vulnerable angle for slope failure initiations (Annual Report of LHMP 2017). The intersects of both “negative values of plan curvature” and “slopes greater than 27°” are represented as the susceptible zones for failure initiations (Fig. 10). For terrain in the study area, plan curvatures with highly negative values indicate greater sideward concave representation leading to the recognition of more morphological units vulnerable to failure propagation. Terrain with plan curvatures of less than −1 (Plan curvature < −1) intersecting with the minimum angle for the failure initiation (27°), represent the highest susceptibility for failure initiations. With this interpretation, two susceptibility zones for future failures are identified (Fig. 10). Using the raster calculator in ArcGIS, the zones with the highest susceptibility for failure initiation are derived by (Plan curvature < −1) × (slope > = 27°). The medium susceptibility zone for failure initiation is defined by (Plan curvature <0) × (slope > =27°) using the raster calculator in ArcGIS.

Fig. 10

Susceptibility map for failure initiation prepared through the geomorphological analysis by slope shading and plan curvature analyses

5 Discussion

The LiDAR data used for this study was obtained prior to the recent Ganthuna Udaganma failure incident date: 17th of May 2016. The LiDAR survey was carried out during the period from December 2015 to April 2016 by JICA (JICA 2016). The susceptibility map prepared using geomorphological analysis (raster plan curvature analysis method) represented zones susceptible to future failures (Fig. 10). As this map was completed before the recent Ganthuna Udagama landslide, the successful prediction of failure initiation (Fig. 10) warranted further inspection of the method’s precision and accuracy. To this end, the boundary of the recent failure is used to determine the best values for the susceptibility evaluation Eq. (1).

Negative values of plan curvature (<0) represent the sideward concave at the pixel, with higher negative values representing greater concavity, and are indicators of the propagation paths. This phenomenon can be observed with plan curvatures < −0.5, plan curvatures < −1, plan curvatures < −1.5, plan curvatures < −2, and more. If the minimum angle for failure initiation of the particular terrain (determined using the geomorphology approach) intersects with these values, then susceptibility zones be confidently mapped. For the terrain under investigation, the minimum slope angle for failure initiation is a fixed value. Thus, to identify terrain susceptible to failure initiation only requires increasing negative values of plan curvatures. The best agreement for the recent failure is (plan curvature <0) × (slope > =27°). Greater precision on the size of zones and locations for failure initiation can be gained by increasing the negative values (e.g., plan curvature < −0.5, plan curvature < −1, plan curvature < −1.5, plan curvature < −2) in Eq. (1). For example, a reduction of the area of susceptible zones obtained from (Plan curvature <0) × (slope > = 27°) is achieved by increasing the negative values (plan curvature < −0.5, plan curvature < −1, plan curvature < −1.5, plan curvature < −2) (Fig. 11).

Fig. 11

Susceptibility evaluation for failure initiation by geomorphological analysis

Field verification confirms the exact locations (1 and 2) of the Ganthuna Udagama failure zones (Fig. 12). The flow path contains transported soil (colluvial deposits) with boulders that may further settle in future (Fig. 12). The map identifies susceptibility zone number 3 as highly vulnerable to failure in future (Fig. 11). This zone is also just upstream of a past landslide indicating susceptibility to reactivation of the past failure. Field observations of large boulders confirm the possibility of failures (Fig. 13).

Fig. 12

Photos taken from field verification process in recent failure (Fig. 11). (a) Location 1, (b) Location 2, (c) Location 6, (d) unstable colluvial deposits in location 7

Fig. 13

Photos taken from unstable zone that taken from field verification process (Fig. 11). (a) Location 8, (b) Location 3

Zones 4 and 5 are also highly susceptible to reactivation-type failures in the future (Fig. 11).

Other high-susceptibility zones (6 to 8) are categorized as highly susceptible to failure as first initiations (Fig. 11).

Predicted susceptibility zones of future failures cover an area of 64,610 m² (Fig. 11) or 19.2% of the total study area. In the center of the study area, the main valley drains from east to west.

Here, most zones identified as susceptible to failure initiation led to the failure of valley walls. Past landslides identified by geomorphological analyses fail into this valley. Recent failure also failed into this valley.

In contrast, the morphology of the south-west part of the study area consists of susceptible zones for failure initiation in valleys trending south-east to north-west.

Villages are the lowest administrative level in Sri Lanka. The study area consists of two villages called “Ganthuna Udagama”, and part of “Narangala” in the south-east. Most of the failure susceptibilities intersect with “Ganthuna Udagama” compared to “Narangala” village.

Identifying the susceptibilities for failure initiation within the lower administrative level is more important to disaster management activities in the future, considering the public safety and socio-economic consequences of catastrophic landslide events.

6 Conclusions

Geomorphological analyses identifying past failures and the potential susceptibility zones for failure initiations were successfully applied to metamorphic terrains in Sri Lanka. Results vary depending on the resolution of the DEM used in ArcGIS. This study demonstrates that LiDAR data with at least 2 m resolution is recommended to accurately define and assess susceptibility zones of future failures. Those zones identified with the highest susceptibility to landslide initiation can be flagged for future disaster management activities.

Zones with the predicted highest susceptibility for failure initiation were field-verified with observations of the recent and past failure boundaries in Ganthuna Udagama.

However, the highest susceptibility zones of the map are not fully field verified or filtered with expert knowledge. Moreover, the susceptibility evaluation is based on current morphological landform units (i.e., the small surface feature) that can be easily changed by natural and anthropogenic activities. Anthropogenic involvement in landform changes must be better understood.

Damage zone assessment and risk mapping can be undertaken by incorporating the results of an evaluation of susceptibility to failure initiation. Susceptibility zones identified are only for failure initiations and do not show all damage zones of initiation, flow paths, and deposition areas. The results presented in Fig. 10 can be used to delimit initiation zones (with maximum potential depths from overburden maps produced by NBRO) to simulate the downslope movement, and to identify the total potential damage zones in the study area. The potential damage zones can then be used to identify all the “elements at risk” and “exposure” (i.e., spatial overlay of geological hazards and elements at risk) to calculate the spatial distribution of landslide risks. Both damage zone assessment maps and risk maps should be used for future landslide disaster management in Sri Lanka.

Lastly, our results can be compared and combined with Landslide Susceptibility Maps produced by the National Building Research Organization obtained by a terrain factor-based model and directly used for community-based future disaster management activities in Sri Lanka.