Multiple Landslides in an Area Draped in Volcanic Matters: The Dual Impacts of Rains and Earthquakes

Abstract

The devastating 2018 Hokkaido Eastern Iburi Earthquake, with a moment magnitude (M_w) of 6.7, inflicted significant damage in the eastern part of Hokkaido, Japan, resulting in the tragic loss of 41 lives, with landslides accounting for 36 fatalities. The mountainous epicentral regions experienced a high number of landslides, primarily induced by the movement of tephra strati originating from eruptions associated with Shikotsu Caldera, Mt. Tarumae, and Mt. Eniwa. The combined intensity of the earthquake and the accumulation of rainfall played crucial roles in initiating these landslides. To gain comprehensive insights into the dynamics of this event, we utilized the Multi-Scale Simulator for the Geo-Environment (MSSG) to simulate the spatial and temporal patterns of rainfall from August 6 to September 5, 2018. Our simulation results were effectively validated against observed data from weather stations operated by the Japan Meteorological Agency (JMA) in the epicentral area. Impressively, the simulation accurately captured the intricate variations in rainfall, taking into account the influence of the region's mountainous terrain. Notably, our findings revealed a noteworthy relationship: as the peak ground acceleration (PGA) increases, the required threshold of rainfall for triggering landslides decreases. This study enhances our understanding of the complexities surrounding landslide dynamics in earthquake-prone areas and contributes to improved disaster preparedness and mitigation efforts.

1 Introduction

Landslides are typically triggered by specific events such as heavy rainfall, intense earthquakes, or a combination of both. It is important to consider the combined effect of these factors, as even a seemingly minor event can ultimately lead to a devastating landslide if the mass is brought close to the tipping point. Recent research by Martino et al. (2022) has revealed a noticeable increase in landslide activity following low-magnitude earthquakes. Additionally, Sassa et al. (2010) conducted an enlightening study using real monitored earthquake records and an undrained dynamic-loading ring shear test. They demonstrated that a minor earthquake with a magnitude of 2.6, 22 km west of the source area, could have triggered the catastrophic 2006 Leyte landslide in the Philippines, claiming the lives of over 1000 people.

On September 6, 2018, at 3:08 a.m. JST, a powerful earthquake with a magnitude of M_w 6.7 struck the northern Japanese island of Hokkaido, leaving a lasting impact. The Japan Meteorological Agency (JMA) officially named it the "2018 Hokkaido Eastern Iburi Earthquake." This seismic event unleashed a series of devastating consequences, including the triggering of over 7,000 landslides within an approximately 20 km × 20 km area near Atsuma Town, as documented by Murakami et al. (2019a, b). These landslides ravaged homes that were sparsely scattered along the foothills of the mountains. The scale of the landslides caused by this intense tremor was unprecedented, encompassing an extensive area of 44 km², as the Japanese Ministry of Land, Infrastructure, Transport, and Tourism (MLIT) reported. This record-breaking extent surpassed any previous occurrence since the Meiji Era (1868-1910), highlighting the magnitude of the disaster. Tragically, around 80% of the 41 victims lost their lives due to suffocation, further emphasizing the severe consequences of this catastrophic event.

The intense seismic activity undeniably played a significant role in triggering the numerous landslides. Some news reports, such as The Asahi Shinbun (2018), raised the possibility that the landslides could have been influenced by the rainfall brought by Typhoon Jebi, which passed through Hokkaido one day before the earthquake, leaving a trail of destruction in its wake. However, the daily rainfall recorded at the Atsuma AMeDAS (Automated Meteorological Data Acquisition System) station, located near the area affected by multiple landslides, was only 12 mm. More importantly, the region had experienced rainfall that was 1.5 times higher than the average for August, adding further context to the weather conditions leading up to the event.

This unfortunate event has provided a unique and valuable opportunity to gain insights into its underlying causes, specifically the role of preceding rainfall and intense seismic activity. This event is particularly significant due to the considerable number of comparable landslides observed in the epicentral area, characterized by the presence of moist pumice-rich volcanic materials. In this article, we delve into the factors contributing to the landslides in the Atsuma region by employing the Multi-Scale Simulator for the Geo-Environment (MSSG) (Takahashi et al. 2006, 2013). This sophisticated model, designed to serve as a coupled non-hydrostatic atmosphere-ocean-land simulation tool, allows us to simulate the rainfall patterns in August and subsequent minor precipitation events, enabling a comprehensive analysis of their impact.

2 Multiple Landslides in the 2018 Hokkaido Eastern Iburi Earthquake

The epicentral area of the 2018 Earthquake in the eastern Iburi region of Hokkaido, located approximately 50 km southeast of Sapporo, was heavily affected by multiple landslides. These landslides were densely distributed throughout the mountainous terrain, with altitudes ranging from 200 to 400 m. The epicentral area is characterized by ridges and valleys, stretching from NNW to SSE. Smaller mountain streams have eroded the primary valley walls within this region, carving deep channels down to lower elevations. Figure 1 illustrates that multiple landslides appeared on these valley walls.

Fig. 1

Landslide inventory map prepared after the 2018 Hokkaido Eastern Iburi Earthquake (Coordinate Reference System: WGS 84 / UTM zone 54 N)

The basement complex of the area primarily consists of sedimentary rocks from the Neogene tertiary system, including layers of mudstone, siltstone, sandstone, and conglomerate (Ozaki and Komatsubara 2014). However, the entire region is significantly covered by three distinct pyroclastic fall deposits resulting from volcanic eruptions west of the area. One such notable feature is the Shikotsu Caldera, a magnificent volcanic caldera formed approximately 40,000 years ago through a massive eruption. Mount Tarumae, one of Hokkaido's most active volcanoes, has a long history of eruptions dating back around 20,000 years. Approximately 9000 years ago, Mount Eniwa volcano experienced a significant eruption that profoundly impacted the surrounding landscape. Consequently, the majority of the multiple landslides in the area were shallow and attributed to these layers of volcanic ash and debris, known as tephra strati (Kawamura et al. 2019).

Utilizing the LiDAR Point Cloud data obtained from the Geospatial Information Authority of Japan (GSI) and the multiple landslides polygons provided by Kita (2018) conducted an estimation revealing that the total area affected by the multiple landslides spans approximately 44 km². In another study, Murakami et al. (2019a) used a 5m-resolution Digital Elevation Model (DEM) provided by the GSI to examine the distribution of average gradients in the Towa River catchment, where shallow landslides accounted for 35.6% of the entire catchment area (refer to Fig. 2). Notably, the analysis indicated that most frequently occurring slope angle (mode value) was 27.5 °degrees, with over 70% of the slope angles measuring below 30 ° (Fig. 2 provides a visual representation of this distribution).

Fig. 2

Distribution of the average gradients of shallow landslides within the Towa River catchment (After Murakami et al. 2019a). The Zonal Statistics Plugin (QGIS) was used to calculate the average gradient for each landslide polygon

3 Preceding Rains

At the time of the earthquake, the epicentral area had four operational Automated Meteorological Data Acquisition System (AMeDAS) stations, which were part of the Japan Meteorologi-cal Agency (JMA) network. These stations were located in Abira (42.81333°N, 141.8283°E), Atsuma (42.73°N, 141.8883°E), Mukawa (42.59°N, 141.9333°E), and Hobetsu (42.76167°N, 142.14333°E), with distances between them ranging from 10 to 20 km (see Fig. 3). Figure 4 displays the cumulative rainfall data recorded by these four stations over a two-month period, from July 1 to September 6, 2018, the day of the earthquake. While the cumulative values differ among the stations, the overall patterns are similar. To emphasize this similarity, Fig. 5 normalizes the data, with the final normalized cumulative values converging to 1.0 on September 6. This normalization highlights that the epicentral area likely experienced a comparable rainfall pattern, albeit with varying amounts of rainfall recorded at each station. Specifically, the pattern consisted of heavy rain in early July lasting for two weeks, followed by three weeks of sunny weather, and then intermittent rainfall episodes from August 4 to September 5.

Fig. 3

AMeDAS Stations by JMA in the Epicentral Area: Atsuma-Horonai Station Commences Observation Three Weeks After Earthquake, Aiding Reconstruction Efforts (Coordinate Reference System: WGS 84 / UTM zone 54 N)

Fig. 4

Cumulative rains at four AMeDAS stations over about two-month from July 1 to September 6, 2018

Fig. 5

Normalized Cumulative Rainfall: Two-Month Observation at Four AMeDAS Stations in the Epicentral Area (July 1 - September 6, 2018)

The influence of rainfall on landslides highly depends on the time-lapse between rainfall and the earthquake. This time-lapse plays a crucial role in determining the moisture conditions within the porous volcanic materials. During rain events, precipitation infiltrates the ground, leading to an increase in moisture content. However, subsequent drainage through seepage flows and evaporation from the ground surface can decrease moisture content. These fluctuations in moisture content directly impact slope stability and can have significant implications for the occurrence of landslides.

At this stage, our primary focus is directed toward the one-month period leading up to the earthquake on September 6. This particular time frame was chosen because the porous pumice, a prevalent material in the region, has a high-water retention capacity. Additionally, as depicted in Figs. 4 and 5, it is plausible that the entire epicentral area experienced a consistent pattern of rainfall. By specifically considering the one-month period of rain, our objective is to investigate the potential impact of this sustained precipitation on the subsequent occurrence of landslides triggered by the earthquake.

4 Method

The MSSG model(Takahashi et al. 2013, 2006), developed through a collaboration between JAMSTEC, Tokyo Institute of Technology, and Waseda University, offers a comprehensive and advanced tool for simulating and analyzing complex weather and climate phenomena. Its high-resolution capabilities, innovative grid systems, and consideration of turbulence-enhanced cloud dynamics make MSSG a valuable asset in understanding atmospheric processes at various scales and investigating rainfall patterns in mountainous regions susceptible to landslides.

For global-scale simulations, MSSG employs the innovative Yin-Yang grid system (Kageyama and Sato 2004), resembling the interconnected skins of a tennis ball. This grid system utilizes square patterns that minimize distortions, effectively overcoming challenges associated with the polar singularity. By accurately representing grid shapes and sizes, MSSG ensures a more precise depiction of global phenomena. Conversely, for regional-scale simulations, MSSG utilizes the conventional latitude-longitude system, offering a reliable framework to analyze localized weather phenomena and their regional impacts.

A notable feature captured by MSSG in its simulations is the turbulence-enhanced collision of cloud droplets. This phenomenon accounts for the rapid growth of cloud droplets, leading to the swift initiation of rainfall during the early stages of cloud development. MSSG explicitly incorporates this effect, which holds particular significance in mountainous regions facing prevailing winds. This characteristic makes MSSG particularly suitable for simulating rainfall in such environments prone to landslides, as it accurately captures the precipitation dynamics.

5 Rainfall Modeling

To conduct the numerical simulation, we employed a multi-domain approach, utilizing three nested domains with progressively finer grid spacings of 8 km, 2 km, and 300 m (refer to Fig. 6). The largest domain covered a significant portion of Hokkaido Island, while the smallest domain was specifically designed to encompass the target area of the epicenter. The geographical boundaries for this innermost domain were defined as 141.777°E to 142.0333°E in longitude and 42.5837°N to 42.83731°N in latitude. In order to ensure the accuracy and reliability of the simulation results, three AMeDAS stations, namely Abira, Atsuma, and Mukawa, were included within this domain for verification purposes.

Fig. 6

Nested domains (Coordinate Reference System: WGS 84 / UTM zone 54 N)

Prominent meteorological agencies worldwide, such as the Japan Meteorological Agency (JMA), the United States National Centers for Environmental Prediction (NCEP), and the European Centre for Medium-Range Weather Forecasts (ECMWF), have developed advanced global numerical weather forecasting schemes. JMA, in particular, has been providing global weather forecasts since 1988, continuously enhancing its model by incorporating data from newly launched satellites and improving resolution in both the vertical and horizontal dimensions. To simulate the 2018 event, the boundary conditions for the outermost domain utilized the Global Spectral Model (GSM2003), featuring a horizontal grid spacing of 20 km and 128 layers extending up to the uppermost layer at 0.1 hPa.

Figure 7 presents the distribution of one-month simulated rainfall within the innermost nested domain, revealing a distinct concentration of orange polygons in the upper-right quarter. These polygons correspond to multiple landslides triggered by the M_w 6.7 earthquake that followed the August rain, with the largest open circle marking the earthquake's epicenter. Surrounding the epicenter, smaller open circles indicate the locations of aftershock epicenters recorded on September 6 and 7. The swarm of landslides exhibits partial overlap with the cluster of aftershock epicenters. Interestingly, even the northwestern section of the landslides, situated several kilometers away from the outer periphery of the aftershock epicenter cluster, aligns with the area of intense rainfall accumulation. Hence, this figure underscores the contribution of both the earthquake's intense shaking and the accumulated rainfall to the occurrence of these numerous landslides.

• (1) Centroids of landslide polygons within the innermost domain for rain simulation.

• (2) One-month cumulative rains and hypocentral distances at centroids of landslide polygons.

In addition, the right side of Fig. 7 includes three sub-figures depicting the rainfall accumulations at the AMeDAS stations (Abira, Atsuma, and Mukawa). The observed rain accumulations are represented by thick orange lines, while the simulated accumulations for the three nested computational domains are also compared. As the grid spacing decreases, the estimated rainfall accumulation is closer aligned to the observed values at each station. This alignment highlights the improved capability of MSSG to capture detailed variations in rainfall patterns influenced by mountainous regions. Although the MSSG simulation did not perfectly replicate every event due to spatial and temporal fluctuations, the one-month simulated rainfall accumulations at the three AMeDAS stations (280 mm, 210 mm, and 200 mm) eventually concur with the observed values (250 mm, 190 mm, and 180 mm), thus validating the effectiveness of the current approach.

Fig. 7

Spatial distribution of simulated one-month rainfall within the finely-resolved nested domain and a comparative analysis of simulated and observed rainfall accumulations at AMeDAS Stations using MSSG

6 Impact of Rains and Earthquake

The peak ground acceleration (PGA) of the bedrock gradually diminishes as the hypocentral distance increases. Hence, in order to assess the influences of these contributing factors on the occurrence of multiple landslides in the region, the hypocentral distances and the values of one-month rainfall accumulation were estimated at the centroids of the landslide polygons. Firstly, the centroids of the 3551 landslide polygons within the innermost computational domain, NEST02, were extracted (see Fig. 8(1)). Subsequently, the hypocentral distances, representing the distances from all these 3551 points to the seismic focus of the main shock (42.6900°N, 142.0067°E, Depth = 37 km), were determined. The raster values of the rainfall accumulation were then sampled at these points, as depicted in Fig. 8(2). The upward-to-the-right inclination of the point cluster in this figure indicates that as the hypocentral distance increases, a higher amount of rainfall is required for a landslide to initiate sliding.

Fig. 8

Simulation of One-month cumulative rains (mm) in the epicentral area. (Coordinate Reference System: JGD2011 / Japan Plane Rectangular CS XII)

To estimate the PGA values at the centroids of the landslide polygons within the innermost domain NEST02, we utilized an empirical attenuation formula proposed by Si and Midorikawa (1999). This formula serves as a reliable method for predicting the PGA values based on specific parameters.

$$ {\log}_{10}A=b-{\log}_{10}{X}_{eq}-k{X}_{eq}, $$

(1)

where

$$ A=\mathrm{peak}\ \mathrm{ground}\ \mathrm{acceleration}\;\left(\mathrm{PGA},\frac{\mathrm{cm}}{{\mathrm{s}}^2}\right)\;\mathrm{or} $$

$$ \mathrm{peak}\ \mathrm{ground}\ \mathrm{velocity}\;\left(\mathrm{PGA},\mathrm{cm}/\mathrm{s}\right) $$

$$ b=a{M}_w+ hD+d+e, $$

(2)

with

a = 0.5, M_w= Moment magnitude, h = 0.0036, D = focal depth (km),

$d=\Big\{{\displaystyle \begin{array}{c}0.00\kern3.1em \mathrm{for}\ \mathrm{crustal}\ \mathrm{earthquakes}\\ {}0.09\kern2.1em \mathrm{for}\ \mathrm{interplate}\ \mathrm{earthquakes}\\ {}0.28\kern0.62em \mathrm{for}\;\mathbf{intraplate}\ \mathbf{earthquakes}\end{array}},\kern0.62em e=0.60$,

and X_eq = Equivalent hypocentral distance (km),

$$ k=\Big\{{\displaystyle \begin{array}{c}0.003\kern0.62em \mathbf{for}\;\mathbf{PGA}\\ {}0.002\kern0.62em \mathrm{for}\;\mathrm{PGV}\end{array}} $$

The equation shown above incorporates several variables, including the moment magnitude (M_w) of the earthquake, the fault distance or equivalent hypocentral distance (X), the focal depth (D), and constants representing the earthquake type (crustal, interplate, or intraplate). The equivalent hypocentral distance (X) is the energy-weighted arithmetic mean of the fault distances from each landslide to the sub-faults forming the complete fault rupture plane. In the case of this earthquake, since most of the aftershocks on September 6 had focal depths deeper than 30 km and close to the 37 km deep hypocenter of the main shock (as depicted in Fig. 9), it can be assumed that the equivalent hypocentral distance (X) is nearly equal to the hypocentral distance. Based on this assumption, Fig. 10 compares the estimated PGA values and the recorded values at 27 observatories in Hokkaido (Japan Meteorological Agency 2018). It is important to consider that not all observatories were situated on bedrock, which explains why the line of the attenuation formula aligns with the lower bound of the group of 27 points representing the observed PGA values. With this validation, Fig. 11 illustrates the estimated PGAs and the values of rainfall accumulation at the centroids of the 3551 landslide polygons within the innermost domain NEST02. The downward-to-the-right inclination of the point cluster in this figure signifies that as the PGA increases, less rainfall is required for a landslide to initiate sliding.

Fig. 9

Probing Depths of Aftershocks: Focal Depths Analysis on September 6, 2018 (The width of each circle represents the magnitude of the corresponding aftershock, which is also shown by the number inside).

Fig. 10

Mapping Ground Shaking: Observation of Peak Ground Accelerations (PGAs) at 27 Hokkaido Observatories and Estimating PGA Attenuation Based on Epicentral Distance

Fig. 11

Simulation of One-month cumulative rains (mm) in the epicentral area. (Coordinate Reference System: JGD2011 / Japan Plane Rectangular CS XII)

7 Conclusions and Future Works

The devastating 2018 Hokkaido Eastern Iburi Earthquake, with a moment magnitude (M_w) of 6.7, inflicted severe damage in the eastern part of Hokkaido, Japan, resulting in the tragic loss of 41 lives, with landslides responsible for 36 of those fatalities. The mountainous epicentral regions experienced numerous landslides, primarily triggered by the movement of tephra strati originating from eruptions associated with Shikotsu Caldera, Mt. Tarumae, and Mt. Eniwa. The combined intensity of the earthquake and the accumulation of rainfall played crucial roles in initiating these landslides.

To gain deeper insights into the dynamics of this event, we utilized the Multi-Scale Simulator for the Geo-Environment (MSSG) to simulate the spatial and temporal patterns of rainfall from August 4 to September 5. Our simulation results were validated against observed data from weather stations operated by JMA in the epicentral area. Impressively, the simulation successfully captured the intricate variations in rainfall influenced by the region's mountainous terrain. Notably, our findings indicated that less rainfall is required for a landslide to be triggered as the PGA increases.

To facilitate further rational discussions, we need to shift focus from one-month cumulative rainfall to soil water indices that quantify the moisture conditions of the soil. However, this process necessitates additional hydrogeological and geotechnical studies, particularly on the porous pumice—a prevalent material in the region known for its high-water retention capacity. Furthermore, slope angles must also be considered, and detailed findings in this regard will be presented in forthcoming publications.

Appendix: Timeline

The events in this article are presented in Japanese Standard Time (JST), which is 9 h ahead of he Coordinated Universal Time (UTC).