Keywords

1 Introduction

Sri Lanka’s landslides susceptibility is primarily attributed to its geological, topographical, and climatic conditions. The island’s mountainous terrain, composed of steep slopes and unstable geological formations in the central highlands, increases landslide vulnerability (Ratnayake and Herath 2005; Jayasinghe et al. 2018). These geological factors and a tropical climate can cause weathering and erosion. Additionally, Sri Lanka’s monsoon climate and high precipitation levels, especially in the central and southern regions, lead to soil saturation and increased pore water pressure, triggering landslides (Ratnayake and Herath 2005). Changes in rainfall patterns in recent years have exacerbated the situation (Jayasinghe et al. 2018). Poor land-use practices, such as deforestation, mining, and agriculture on steep slopes, further weaken the soil structure, making landslides more frequent and damaging, affecting infrastructure, property, and lives (Nissanka et al. 2015).

The Athwelthota landslide is one of the recent examples of a devasting rain-induced rapid and long-travelling landslide (RRLL) in Sri Lanka. It occurred on May 26th, 2017, in the Palindanuwara Divisional Secretariat of the Kalutara District (Fig. 1). This landslide, triggered by unprecedented daily rainfall of more than 241 mm, resulted in nine deaths and severe damage to infrastructure, including the destruction of seven houses. The Kalawana-Baduraliya Road and Palan Ganga (River) were also blocked with debris, and the Athwelthota Gangaramaya temple was partially destroyed.

Fig. 1
A. A satellite image of Sri Lanka presents the location of the Athwelthota landslide. B. A satellite image presents a close-up view of the location of the Athwelthota landslide. C. A drone image of the Athwelthota presents the depositional area, flow path, landslide crown, and others.

Satellite (a and b) and drone image (c) of Athwelthota landslide location including its main features and sample locations. (Source base map: Google satellite imagery)

Full size image

Similar to the above event, the island nation experiences recurring landslide events. Hence, it became a contemporary need to minimize the catastrophic loss due to landslides. In this regard, a joint research program was initiated between the International Consortium on Landslide (ICL) and the National Building Research Organisation (NBRO). The title of the project is “Development of Early Warning Technology of Rain-Induced Rapid and Long-Travelling Landslides in Sri Lanka (Project RRLL)” (Konagai et al. 2021). One of the vital objectives of the project RRLL is to identify the causes and contributing factors of the landslide that happened in Athwelthota, one of the pilot sites of this research study.

This study aims to reproduce the 2017 Athwelthota landslide and discuss future hazards of nearby potential landslides. For this aim, the study will use a combination of historical records, geomorphological field surveys, geospatial data, soil parameters, rainfall data, geospatial information, landslide geometry, and other relevant data.

2 Methodology

This study employs a multi-pronged approach to investigate the Athwelthota landslide hazard. The methodology consists of four key components:

  1. 1.

    Desk study: The first step was reviewing the available literature and geospatial data, including satellite imagery. This review helped identify potential landslide areas and gather information on past landslides in the region.

  2. 2.

    Fieldwork: The fieldwork was conducted to identify past and potential landslides in the study area. This work involved geomorphological mapping and the collection of soil samples for laboratory testing. A few locations were selected for soil sampling (Fig. 1c) by considering the soil type variation of the past landslide and potential landslide area.

  3. 3.

    Geotechnical laboratory testing: Soil samples were tested at the Geotechnical Laboratory of NBRO. The primary objective of this testing was to obtain the geotechnical parameters required for numerical simulation. The undrained dynamic ring shear loading apparatus (ICL-2) was used to determine the shear strength properties of the soil (Sassa et al. 2004a).

  4. 4.

    Numerical analysis using LS-RAPID: The LS-RAPID (Landslide Simulator for Risk Assessment and Prediction of Inundation Disasters) model, developed by Sassa et al. (2010), has emerged as an essential tool in the field of landslide research and risk assessment. The foundation of the LS-RAPID model can be traced back to Sassa’s work in 1988 when he initially formulated a geotechnical model for landslide motion. The laboratory test results, field observations, and geospatial data were inputs to the LS-RAPID numerical model to simulate the past landslide in 2017. The LS-RAPID model has been successfully used to simulate landslides in various geological and environmental conditions, providing valuable insights into landslide mechanics and failure factors. For example, the model was used to simulate the Aranayake landslide in Sri Lanka (Tan et al., 2020), the Oso landslide in the United States (Brien et al. 2016), and the Izu-Oshima landslide in Japan (Tsuyoshi et al. 2017).

3 Results and Discussion

Satellite imagery shows that the study area comprises several undulations, which could be possible past landslides. However, field observations revealed two past landslides that had failed earlier. One of these landslides occurred in 2017, and the other occurred at an unknown period (Fig. 2). Both landslides shared the same flow path and depositional area but had separate initiation areas. Both these initiation areas are located near the valley’s axis, which trends north.

Fig. 2
A drone image of the Athwelthota. It presents the initiation area of landslide occurred in 2017, depositional area, old landslide scar, tension cracks, and others.

Identified Past and Potential Landslides and their features in Athwelthota

Full size image

Two potential landslides were identified during the field survey on the slope to the right of the valley, next to the previously identified past landslides (Fig. 2). These landslides exhibited tensional cracks and depressions as potential landslide features. These tensional cracks occurred during the same period as the failure of the past landslide in 2017.

After the sample collection phase, laboratory tests were performed, and the results of these tests and their corresponding parameter values from the ring-shear apparatus are presented in Table 1.

Table 1 Summary of undrained ring shear test results
Full size table

Rain infiltration leads to an increase in the groundwater level, thus generating significant pore pressure to trigger an RRLL. The pore pressure ratio Ru, a dimensionless parameter defined as the ratio of the pore water pressure to the vertical effective stress (Bishop and Morgenstern 1960), rationally provides the RRLL’s threshold. The SLIDE model (Liao et al. 2010) was used in this study to calculate the Ru value.

The chronological record of the rain in May 2017 (Fig. 3(b)), together with the landslide morphology, soil properties, and infiltration rates, was given as input to the SLIDE model to obtain the buildup of the Ru value (Fig. 3(a)). The laboratory testing results and estimated geotechnical parameters in Table 2 were used to simulate the Athwelthota landslide. The generated results for the Athwelthota landslide mass are depicted in Fig. 4.

Fig. 3
2 graphs. Top. A line graph of the pore pressure ratio plots an ascending line between 0.0 to 0.5 R u on the y-axis and 0 hours 0 minutes to 85 hours on the x-axis. Bottom. A bar graph presents 30 minutes of rainfall. It depicts the bars with fluctuating trends.

(a) Calculated Ru values, (b) rainfall histogram in 24th–26th May, 2017

Full size image
Table 2 Geotechnical parameters utilized in the LS-RAPID software for Athwelthota Landslide
Full size table
Fig. 4
A to C. 3 drone images present the Athwelthota landslide indicating calculated sliding mass distribution area and mass expression the simulation results. On the right side, there are corresponding monitor values with elapsed time and 2 graphs of pore pressure ratio and 30 minutes rainfall.

Simulated landslide failure process for the Athwelthota landslide at each stage using the LS-RAPID with Ru and rainfall data: (a) initiation, (b) start of sliding, and (c) end of sliding

Full size image

At 67 h 01 min in the simulation, when the pore pressure ratio (Ru) reached 0.43, the central part of the soil mass parching atop the slope started to detach (Fig. 4(a)). Subsequently, at approximately 68 h and 36 min (Ru = 0.43), the entire landslide mass started sliding (stage 2 in Fig. 4). Finally, the fluidized soil mass came to a halt at 68 h and 38 min after the simulation started (stage 3 in Fig. 4), which corresponds to 4:38 a.m. on May 26th, 2017 (Table 3). The actual landslide was considered to have occurred at around 5:15 a.m. on that day (NBRO n.d.), which was eventually close to 4:38 a.m., the time estimated by using LS-Rapid. Although a relatively small landslide occurred as a branch of the main landslide a few hours later, only the main landslide of the Athwelthota Landslide was considered for the numerical modelling due to its complexity.

Table 3 Model and real time of Athwelthota landslide
Full size table

Since the LS-RAPID model with the given parameters could reproduce the real 2017 RRLL adequately, as described above, it was then used to simulate the behavior of the identified two potential failures, L1 and L2 (Figs. 5 and 6), and assess their potential impacts. The morphology of the potential landslides was inferred from the local geomorphology and geological observations, with particular attention to the features of previous landslides.

Fig. 5
2 drone images of the Athwelthota landslide location exhibit the calculated sliding mass distribution area and mass expression the simulation results. It presents the corresponding monitor value with elapsed time and a graph of the pore pressure ratio. The graph presents R u value = 0.67.

Simulated landslide failure process for potential landslide L1 using the LS-RAPID with Ru data (a) initiation and (b) failure

Full size image
Fig. 6
2 drone images of the Athwelthota landslide location indicate the calculated sliding mass distribution area and mass expression the simulation results. It includes the corresponding monitor value with elapsed time and a graph of the pore pressure ratio. The graph presents R u value = 0.64.

Simulated landslide failure process for potential landslide L2 using the LS-RAPID with Ru data (a) initiation and (b) failure

Full size image

A usual simulation approach, which incorporates porewater pressure as a critical parameter within the model, was employed to simulate these potential failures. It is important to mention that these potential landslides were not exposed to a triggering rainfall event. Therefore, the simulation was conducted, increasing values of the Ru parameter until the landslide mass reached the critical point of equilibrium and started sliding. The two hypothetical landslides, L1 and L2, have Ru thresholds of 0.67 and 0.64, respectively. The notable aspect here is that these Ru values surpass the corresponding Ru value of the past landslide. This disparity in Ru values illuminates why the prospective landslides had not yet been set into motion. The higher Ru values indicate that the slopes have been more stable than those that failed in the past, which included the 2017 landslide. Though the potential landslides appear to be relatively more minor than the past landslides, they still pose a significant risk of debris flow and long traveling, damaging the surrounding environment and infrastructure while potentially endangering lives and properties.

4 Conclusions

This study focused on applying LS-RAPID numerical simulation software to understand the future landslide potential around the Athwelthota past landslide. Firstly, the numerical model was validated by reproducing the landslide in May 2017 by incorporating geospatial data and geotechnical parameters determined using the ICL-2 ring shear apparatus. It was found that the time of landslide initiation and the landslide boundary reached comparable outcomes with field data recorded. Upon validation, the simulation was extended to assess the probability of two future landslides adjacent to the past landslide. The simulation results found that the Ru value needed to trigger future landslides was higher than the previous landslide. This evidence provides a rationale for not initiating the landslide during the May 2017 rain spell.

However, it is crucial to acknowledge the limitations of this study, including data availability, data quality, uncertainties related to model parameters, and variations in soil properties.

Future research endeavors should explore site-specific factors contributing to landslide susceptibility, encompassing land-use practices, vegetation cover, and groundwater conditions.

In summary, this study provides valuable insights into Athwelthota specific landslide dynamics and contributes to the broader knowledge of landslide risk management. By addressing the unique challenges of this region and highlighting the potential for early warning systems and site-specific mitigation strategies, our research underscores the critical role that comprehensive analysis and modeling play in minimizing the catastrophic impact of landslides. This study stands as a testament to the ongoing efforts to safeguard lives, infrastructure, and ecosystems in landslide-prone areas.