case study about landslide

• Unit 4: Anatomy of a tragic slide: Oso Landslide case study

Sarah Hall (College of the Atlantic) Becca Walker (Mt. San Antonio College)

Learn more »

Science and Engineering Practices

Obtaining, Evaluating, and Communicating Information: Compare, integrate and evaluate sources of information presented in different media or formats (e.g., visually, quantitatively) as well as in words in order to address a scientific question or solve a problem. HS-P8.2:

Constructing Explanations and Designing Solutions: Apply scientific ideas, principles, and/or evidence to provide an explanation of phenomena and solve design problems, taking into account possible unanticipated effects. HS-P6.3:

Analyzing and Interpreting Data: Analyze data using tools, technologies, and/or models (e.g., computational, mathematical) in order to make valid and reliable scientific claims or determine an optimal design solution. HS-P4.1:

Cross Cutting Concepts

Stability and Change: Change and rates of change can be quantified and modeled over very short or very long periods of time. Some system changes are irreversible. HS-C7.2:

Patterns: Empirical evidence is needed to identify patterns. HS-C1.5:

Cause and effect: Cause and effect relationships can be suggested and predicted for complex natural and human designed systems by examining what is known about smaller scale mechanisms within the system. HS-C2.2:

Disciplinary Core Ideas

The Roles of Water in Earth's Surface Processes: The abundance of liquid water on Earth’s surface and its unique combination of physical and chemical properties are central to the planet’s dynamics. These properties include water’s exceptional capacity to absorb, store, and release large amounts of energy, transmit sunlight, expand upon freezing, dissolve and transport materials, and lower the viscosities and melting points of rocks. HS-ESS2.C1:

Plate Tectonics and Large-Scale System Interactions: Plate tectonics is the unifying theory that explains the past and current movements of the rocks at Earth’s surface and provides a framework for understanding its geologic history. Plate movements are responsible for most continental and ocean-floor features and for the distribution of most rocks and minerals within Earth’s crust. HS-ESS2.B2:

Performance Expectations

Earth and Human Activity: Construct an explanation based on evidence for how the availability of natural resources, occurrence of natural hazards, and changes in climate have influenced human activity. HS-ESS3-1:

This activity was selected for the On the Cutting Edge Reviewed Teaching Collection

This activity has received positive reviews in a peer review process involving five review categories. The five categories included in the process are

For more information about the peer review process itself, please see https://serc.carleton.edu/teachearth/activity_review.html .

Landslides can have profound societal consequences, such as did the slide that occurred near Oso, Washington in 2014. Forty-three people were killed and entire rural neighborhood was destroyed. In this unit, students consider the larger-scale tectonic and climatic setting for the landslide and subsequently use lidar and SRTM (Shuttle Radar Topography Mission) hillshade images, topographic maps, and InSAR (interferometric synthetic aperture radar) to determine relationships between landscape characteristics and different types of mass-wasting events. They conclude by considering the societal costs of such a disaster and ways that communities in similar situations may mitigate their risk.

Expand for more detail and links to related resources

Activity Classification and Connections to Related Resources Collapse

Grade level.

Learning Goals

Unit 4 learning outcomes × div[id^='image-'] {position:static}div[id^='image-'] div.hover{position:static} a view from a washington army national guard helicopter shows trees that were at the very top of the hill when the mudslide occurred now lie across the slide's width like a woven belt. provenance: photo by sgt. matthew sissel joint forces headquarters, washington national guard https://www.dvidshub.net/image/1209685/oso-mudslide reuse: this item is in the public domain and maybe reused freely without restriction..

• Students will identify several environmental characteristics of a natural landscape that increase an area's mass-wasting susceptibility.
• Students will create a map with appropriate symbology illustrating the morphological characteristics of a mass-wasting event.
• Students will use geodetic data to compare the appearance, existence, and morphology of surface features before and after a mass-wasting event.

Unit 4 Teaching Objectives

• Cognitive: Provide opportunities for students to use climate and geodetic data to characterize the evolution of landscape features after a mass-wasting event has occurred.
• Behavioral: Facilitate student ability to qualitatively assess and map the changes in geomorphic features before and after a mass-wasting event.
• Affective: Facilitate student ability to analyze the societal impact of mass-wasting events.

Context for Use

The content in Unit 4 is appropriate for introductory geology, hazards, and other geoscience courses; sophomore-level courses in which geodesy and/or geomorphology/surface processes are being introduced; or non-geoscience courses where infrastructure planning, Earth hazards, and/or the nature and methods of science are being investigated. Unit 4 activities can be adapted to serve small or large-enrollment classes and can be executed in lecture and lab settings as an in-class activity in which students work in small groups, a lab exercise, or as part of a ~2 week investigation of the use of geodesy to understand surface-process hazards and decision making using the entire Surface Process Hazards module.

This unit precedes Unit 5: Mitigating future disasters—developing a mass-wasting hazard map in which students create a hazard map and report on an area susceptible to mass wasting. If the entire two-week module will not be used, we recommend pairing Unit 4 with Unit 5 to give students the opportunity to use geodetic data to identify landscape features associated with a slow-slip mass-wasting event and classify low, medium, and high-risk areas based on landscape characteristics. However, Unit 4 does build on skills from Unit 2: Reading the landscape and Unit 3: Understanding landslide factors so you may need to have to give some background practice or information on these topics.

Description and Teaching Materials

Part 1: preparatory activity.

• In each of the five study sites that students have looked at in Units 2 and 3, there is a significant mass wastage somewhere in the map area. In order to prepare for Unit 4, have students prepare five-minute presentations on one those mass-wastage events (except the Northern Washington one, as the rest of Unit 4 is about the Oso Landslide in that region). If you have a large class, you may want to have the students work in groups or randomly select one person/team from each of the four sites to present. This process helps students understand more about the societal cost of mass wasting and the different types of ones that can occur.
• Unit 4 preparation student exercise (Microsoft Word 2007 (.docx) 304kB Jun20 17)
• USGS: History of Landslides at the Base of Bare Mountain, Tully Valley, Onondaga County, New York Report
• USGS: Landslide Susceptibility in the Tully Valley Area, Finger Lakes Region, New York
• USGS: Report of the Thistle Slide
• Utah Geological Survey: Geosights: Thistle Landslide Revisited
• USGS: Rockslide-debris Avalanche of May18, 1980, Mt St Helens Volcano
• USGS: Mt St Helens 1980 Cataclysmic Eruption
• AK Dept of Fish & Game: [link https://pubs.er.usgs.gov/publication/70168586 Part 2: Setting the stage 'A submarine landslide source for the devastating 1964 Chenega tsunami, southern Alaska']
• Earthquake, Landslide and Tsunami Hazards in the Port Valdez area, Alaska

In order to place the Oso Landslide in a larger geologic context, students will review the environmental and built characteristics of the northwest Washington locality using the maps from Units 2 and 3.

A) Plate tectonic setting Seismic and volcanic hazards exist along all of eastern Washington state related to the subduction of the Juan de Fuca Plate beneath the North American Plate. This is a good opportunity to relate the plate tectonic setting to the landscape features (faults, linear ridges/streams, mountains) and present hazards. Use the same map-set for Northern Washington as you did in Unit 3. Northern Washington Maps Unit 3-4 (Acrobat (PDF) 41.3MB Oct22 20)

We suggest one of the following already-existing plate tectonics activities:

• GPS Data and Earthquake Hazards in Cascadia - Students learn to read GPS timeseries to calculate velocities and then analyze what the changing velocity across Cascadia means for earthquake hazard.
• Analyzing Plate Motion Using EarthScope GPS Data - Student use a spreadsheet of GPS data from the Plate Boundary Observatory (PBO) to measure the motion of GPS stations in the Pacific Northwest. They generate and map annual velocity vectors of GPS stations.

B) Climatic setting Washington state has an excellent example of high-precipitation region and a rain shadow. High seasonal precipitation focused on the western slopes of the Cascades and arid conditions prevailing over eastern Washington. In this activity students looks at climate data from near Oso, Washington, as well as a precipitation map of Washington to learn more about the general climate of the area as well as in the time leading up to the slide itself.

Washington precipitation map (Acrobat (PDF) 1.2MB Feb25 16) Unit 4: Oso Washington climate assessment student exercise (Microsoft Word 2007 (.docx) 426kB Jun20 17) Unit 4: Oso Washington climate assessment student exercise PDF (Acrobat (PDF) 361kB Aug28 17) Unit 4: Oso Washington climate assessment student exercise answer key -- private instructor-only file Hide Unit 4: Oso Washington climate assessment student exercise answer key This file is only accessible to verified educators. If you are a teacher or faculty member and would like access to this file please enter your email address to be verified as belonging to an educator. Email Adress Submit Unit 4: Oso Washington climate assessment instructor calculations -- private instructor-only file Hide Unit 4: Oso Washington climate assessment instructor calculations This file is only accessible to verified educators. If you are a teacher or faculty member and would like access to this file please enter your email address to be verified as belonging to an educator. Email Adress Submit

Part 3. Mass-wasting anatomy—Oso, WA, Case Study

Along with the assignment below and the map set they have already used, students should be given a lidar hillshade image onto which they will identify, map, and label morphological features of the mass-wasting event. They will answer qualitative questions and will color code their maps to show regions of high/moderate/low risk for damage due to slip. Students will then be shown spatial data of the region at various timesteps (including SRTM [Shuttle Radar Topography Mission], lidar, and InSAR maps of the locality) and answer questions related to the geology and societal impacts of the event.

• Northern Washington Maps Unit 3-4 (Acrobat (PDF) 41.3MB Oct22 20)
• Unit 4 Oso Landslide Exercise (Microsoft Word 2007 (.docx) 34MB Oct21 20) Unit 4 Oso landslide exercise PDF (Acrobat (PDF) 16.4MB Oct21 20)

This file is only accessible to verified educators. If you are a teacher or faculty member and would like access to this file please enter your email address to be verified as belonging to an educator. Email Adress Submit

Teaching Notes and Tips

• This unit seems to go forward in a pretty straightforward way. Although it takes a little extra time, we really do recommend the brief student presentations in Part 1. If students have never done PowerPoint presentations (or other presentation software), you may need to spend a little time explaining what to do or give guidance on what each slide should contain.

Formative Assessment

Most likely, formative assessment will be gathered through discussion with students and observation of students as they work through the exercise. By monitoring classroom activity, the instructor can gauge how students are progressing toward the learning outcomes.

Summative Assessment

The primary summative assessment for the module is the Oso Landslide Exercise. In addition, in order to assess whether students have achieved a sufficient understanding of the societal side of the unit, one of the following questions could be used on a test.

Level 2: Besides the mass of material that buried the town, the Oso mass-wasting event left a mark on the geographic and societal landscape both locally and regionally. Students will be asked to answer the following questions:

• Name and describe two examples of a local indirect effect of the Oso mass-wasting event—near to the town of Oso.
• Name and describe two examples of a regional indirect effect of the Oso mass-wasting event—downstream of the event?

Level 3: Students will research what has been done to plan for the future in the aftermath of the Oso event. They will prepare a one-page narrative describing how the community and/or state is working to prevent or preparing for the future in this landslide-prone region.

References and Resources

• Jin-Woo Kim, Zhong Lu, Feifei Qu & Xie Hu (2015) Pre-2014 mudslides at Oso revealed by InSAR and multi-source DEM analysis , Geomatics, Natural Hazards and Risk, 6:3, 184-194, DOI: 10.1080/1947705.2015.1016556
• Haugerud, R. A. (2014) Preliminary Interpretation of Pre-2014 Landslide Deposits in the Vicinity of Oso, Washington , USGS Open-File Report 2014–1065
• USGS Follow-up March 2015: One Year Later—The Oso Landslide in Washington .
• Wikipedia Oso Mudslide
• Introduction to Geodesy and Surface Process Hazards Presentation (PowerPoint 2007 (.pptx) 14.7MB Jul11 17)
• https://www.oregongeology.org/Landslide/Landslidehome.htm
• http://www.seattle.gov/dpd/aboutus/whoweare/emergencymanagement/
• http://www.portlandoregon.gov/pbem/53935
• http://landslides.usgs.gov/
• http://landslides.usgs.gov/learn/prepare.php
• http://pubs.usgs.gov/circ/1325/

« Previous Page       Next Page »

• Module Overview
• Unit 1: Slip-sliding away: case study landslides in Italy and Peru
• Unit 2: Reading the landscape
• Unit 3: Understanding landslide factors
• Unit 5: Mitigating future disasters: developing a mass-wasting hazard map
• Student Materials
• Instructor Stories
• Join the Community

Already used some of these materials in a course? Let us know and join the discussion »

Considering using these materials with your students? Get advice for using GETSI modules in your courses » Get pointers and learn about how it's working for your peers in their classrooms »

• About this Project
• Accessibility

We encourage the reuse and dissemination of the material on this site for noncommercial purposes as long as attribution to the original material on the GETSI site is retained.

Citing and Terms of Use

Material on this page is offered under a Creative Commons license unless otherwise noted below.

Show terms of use for text on this page »

Show terms of use for media on this page »

• Last Modified: April 09, 2024
• Short URL: https://serc.carleton.edu/189147 What's This?

• Case Report
• Open access
• Published: 24 August 2022

A community-operated landslide early warning approach: Myanmar case study

• Neil Dixon 1 ,
• Alister Smith 1 &
• Matthew Pietz 2  

Geoenvironmental Disasters volume  9 , Article number:  18 ( 2022 ) Cite this article

3789 Accesses

4 Altmetric

Metrics details

A landslide early warning system based on monitoring acoustic emission (AE) generated by slope movements has been developed that can deliver alerts direct to a community at risk, with relevance to low- and middle-income countries. The Community Slope SAFE (Sensors for Acoustic Failure Early-warning) (CSS) approach uses steel waveguides driven into the slope to transmit detected high frequency noise (AE) to a sensor at the ground surface. CSS gives a measure of slope displacement rate. Continuously measured AE is compared to a pre-defined trigger level that is indicative of decreasing slope stability (i.e., landslide initiation), and a visual and audible alert automatically generated so that a community can follow a pre-defined course of action (e.g., evacuation). This paper describes the CSS approach and details a field trial of the system at two sites in Hakha, Chin Sate, Myanmar. The trial, which included training a group of youth Landslide Response Volunteers to install and operate the CSS system, increased landslide awareness and knowledge in the Hakha community, delivered the required real-time continuous operation, and demonstrated the practicality of using the CSS system for community landslide protection.

Introduction

Problem and requirements.

Rainfall-induced landslides cause tens of thousands of deaths annually. They also damage critical infrastructure, impacting on quality of life and costing millions of dollars to repair (Froude & Petley 2018 ). Climate change, coupled with growing populations and urbanisation, is increasing the prevalence of landsliding and their impacts (Kjekstad & Highland 2009 ). Despite the availability of well-established monitoring approaches (e.g., Stähli et al. 2015 ), communities experiencing the largest impacts from landslides are often amongst the poorest and include those that do not have access to bespoke monitoring solutions (e.g., Liu et al. 2016 ; de Assis Dias et al. 2020 ). UNISDR ( 2007 , 2015 ) has called for development of early warning systems (EWS) that can be deployed in low-income economies. EWS have been defined by UNISDR ( 2009 ) as ‘ the set of capacities needed to generate and disseminate timely and meaningful warning information to enable individuals, communities and organizations threatened by a hazard to prepare and to act appropriately and in sufficient time to reduce the possibility of harm or loss’ .

EWS can be classified as alarm , warning and forecasting systems (Stähli et al. 2015 ). This paper focusses on an alarm system to provide alerts direct to a community in the immediate vicinity of a landslide. The goal is to provide sufficient time for implementation of a pre-determined action plan to protect people at risk (e.g., evacuation). This alarm approach is distinct to a warning system whereby experts analyse a situation, or a forecasting system whereby regional-scale danger levels are produced by experts (e.g., based on rainfall thresholds; Nam and Wang 2020 ; Ngandam Mfondoum et al. 2021 ; Kalubowila et al. 2021 ). As noted by Dixon et al. ( 2018 ) amongst others, a landslide alarm EWS for use by vulnerable communities in low- and middle-income countries must fulfil the following criteria:

be affordable (i.e., sufficiently low cost)

easy to install and use

operate in a range of site conditions

monitor at appropriate spatio-temporal resolutions

quantify slope deformations (rates) that can pose a risk

be self-sustaining and require minimal human intervention

transfer information direct to the user (i.e., the alert)

operate in real-time; and

be robust (i.e., minimal false alerts).

It must also be recognised (UNEP 2012 ) that an EWS is not solely about technology, and must comprise all four following elements:

A comprehensive assessment of the risks

A sensor-based monitoring and warning system

A plan for the dissemination of alerts

A strategy for the response of the people at risk

It can be concluded that there is an urgent need for affordable sensor-based EWS that can be operated by communities in low- and middle-income countries.

Landslide early warning approach

Research has established acoustic emission (AE) slope monitoring as a viable alternative to traditional deformation-based measurement techniques (Berg et al . 2018; Dixon et al. 2015a , 2015b ; Smith and Dixon 2015 ; Smith et al. 2014 , 2017a , b ). It uses detection and quantification of super-audible noise (i.e., high frequency) generated by particle-to-particle contacts during deformation of soil in a failure event to derive slope displacement rates. AE can be used to provide an early warning of slope instability by detecting both the development of shear surfaces and accelerating deformation behaviour (Chichibu et al. 1989 ; Fujiwara et al. 1999 ; Koerner et al. 1981 ; Michlmayr et al. 2017 ; Nakajima et al. 1991 ; Smith et al. 2017a ). This significant body of research has proven AE rates are indicative of slope displacement rates, and hence, AE instrumentation can be defined as a slope displacement rate sensor. Activity status of a slope can be obtained (e.g., stable, accelerating and decelerating trends). Also, critical AE rate thresholds (i.e., related to slope velocities) can be established to trigger alerts. Dixon et al. ( 2018 ) detail development and laboratory testing of an AE monitoring approach for use by communities called Community Slope SAFE (Sensors for Acoustic Failure Early-warning) (CSS). A key motivation for developing the CSS approach was to establish a cost effective, subsurface monitoring system that can provide alerts of movements continuously and in real-time direct to those in danger. The expectation is that following training, non-specialist community members can install and set up the system, maintain it and use the EWS. The CSS approach used and evaluated in the Myanmar case study is described in “ Community slope SAFE (CSS) monitoring system ” section.

Myanmar case study

In June 2015, parts of Myanmar were devastated by cyclone Komen and Chin State was one of the most severely affected, with housing and infrastructure damaged and destroyed. Even before the disaster, Chin State was the poorest region of the country and is geographically isolated. Although a mountainous sparsely populated region (Fig.  1 ), it is estimated that over 20,000 people were temporarily displaced in Chin State by extensive flooding and landsliding caused by the cyclone (Cuai 2017 ). Reasons for the high number of people impacted include: higher than expected rainfall (potentially exacerbated by climate change), changes in vegetation cover (including deforestation), unplanned development, an absence of regulations, and a lack of experience and professionals to provide guidance for construction activities. After the event, due to concern of potential future impacts from landslides, 4,000 people from Hakha were moved to temporary displacement camps and consideration was given to relocate Hakha, the state capital with a population of 50,000.

a Diagram of Myanmar showing the location of Chin State (the black circle shows the location of Hakha), and b photograph of Hakha

In a response to this event, Family Health International (FHI) 360, an international not-for-profit organization with offices in Myanmar working to improve the health and well-being of its people, decided that it should be part of their mission to help protect communities from landslides. In 2016, they instigated a search for innovative technological solutions and potential partners to deliver landslide EWS for communities in remote areas of Myanmar. This led to a collaboration between FHI 360 and Loughborough University (LU) to develop a monitoring approach that vulnerable communities can afford and use. From inception, it was considered essential that the end-user community should be engaged with the maintenance and use of the system, and therefore the interface between technology and users was of critical importance. With funding from FHI 360, the authors undertook a trial of the CSS approach at two sites in Hakha.

The research reported in this paper was designed to have an impact on the landslide resilience of communities in Chin State, Myanmar, by:

increasing landslide awareness and knowledge in the community via a trial of the CSS approach

training a group of youth Landslide Response Volunteers (LRV) to install and operate the CSS alarm system and to act as ambassadors within the community; and

evaluating the performance and practicality of using the CSS specifically, and EWS in general, for community landslide protection.

The paper describes the CSS system and details the Myanmar case study. Performance of the system and lessons learned are also addressed. The original contributions of this study include the installation and trial of the landslide EWS in the field environment and operation by the community in Myanmar.

Community slope SAFE (CSS) monitoring system

Figure  2 a) shows a schematic of the CSS system used in this trial, Fig.  2 d) details the main components and operating architecture, Fig.  2 b) shows photographs of the sensor and base station components and Fig.  2 c) shows the method of sensor connection on a waveguide. A full description of the CSS approach, including details of a laboratory study to validate operation of the system and interpret detected AE, is given by Dixon et al. ( 2018 ), with a summary provided below.

a Community Slope SAFE (CSS) AE monitoring approach incorporating a driven waveguide, b CSS sensor and base station components, c a view of a sensor located on the waveguide inside a cover, and d system architecture of the Community Slope SAFE approach (modified after Dixon et al. 2018 )

The AE sensor is attached to a steel tube waveguide driven into the slope (Fig.  2 a, c). The waveguide both intercepts AE generated by soil deformation if the slope starts to move (it can also generate AE as the tube is deformed as described below) and it also provides a low attenuation propagation pathway for AE produced in the slope to travel to the ground surface where they can be detected and quantified by the sensor. Use of driven waveguides simplifies the installation process by using readily available low-cost equipment (e.g., hand operated post rammers), and this also reduces costs, although the depths to which they can be driven depends on the strength and stiffness properties of the host soil. Installation of a waveguide by driving, results in the steel tube being in intimate contact with the in-situ ground forming the slope, which ensures it is sensitive to slope deformations by minimising AE transmission losses at the soil/tube boundary. ‘Noisy’ material is placed inside the waveguide as proposed by Nakajima et al. ( 1991 ) such that as the tube is deformed, straining of the infill also generate AE. Sand is an ideal material to use as the ‘noisy’ infill material as it is cheap and universally available, and it was used in this study. The CSS system comprises the following functions (Fig.  2 d):

Conversion of AE (i.e., stress waves) propagating along the waveguide to voltages using a piezoelectric transducer

Signal amplification and filtering to remove noise at both low (< 15 kHz) and high (> 40 kHz) frequencies that can be generated by the electronics and/or environment

calculation of the signal RMS (root mean square, a measure of signal energy)

averaging AE activity over a defined period (e.g., 30 s) by aggregating the RMS values

comparison of the AE RMS rate over the aggregation period with a pre-determined threshold value; and

generation of an alert message to the community via the base station if the threshold is exceeded, which sets off an audible and visual alert .

A community can use the alert to initiate a pre-agreed action plan such as evacuation of a specified area using agreed routes, inspecting the slope if safe to do so and informing nominated authorities and professionals.

Series of physical model experiments were conducted to evaluate performance of the CSS approach by comparing applied deformation behaviour of active waveguides with the measured AE response. The waveguides were subjected to accelerating deformations to replicate known behaviour in first-time landslides, which accelerate as progressive failure occurs and post-peak strengths are mobilised. Figure  3 a) shows the experimental setup, and full details of the test procedure are reported in Smith et al. ( 2017a ) and Dixon et al. ( 2018 ). The large-scale direct shear device has a central column filled with stiff clay. An active waveguide and a ShapeAccelArray (SAA) in-place inclinometer were installed through the clay column. Displacement of the top block relative to the bottom block was by a pulley system controlled by a hydraulic actuator, which generated shearing in the clay column. The excellent agreement between AE and displacement trends (i.e. cumulative and rate) shown in Figs.  3 b) and c) provides conclusive evidence that AE rates are indicative of ‘slope’ displacement rates, and hence that AE monitoring can be used to provide information on the stability status of a slope.

Large-scale slope failure physical model used to accelerate the top box relative to the bottom and to compare displacement behaviour with AE measured using the CSS system: a annotated photograph of the apparatus; b CSS cumulative AE and SAA measured displacement time series; and c CSS AE rate and SAA measured velocity time series (modified after Smith et al. 2017a and Dixon et al. 2018 )

CSS trial permission

Following extensive discussions, permission was obtained in 2017 from the Director General of the Department for Disaster Response and Resettlement to undertake a trial of the CSS landslide early warning system in Chin State.

Community engagement strategy

A critical element of project delivery was the need to establish collaboration with a local Chin State organisation to lead engagement with the community and hence facilitate delivery of the landslide monitoring program and associated training. An agreement was made between FHI 360, LU and the local collaboration partnership Chin Committee for Emergency Response and Rehabilitation (CCERR). CCERR is a community organization that was already coordinating relief efforts and supporting communities to recover from the 2015 cyclone event and was working to build capacity and resilience to future events. With assistance from CCERR, agreement and support were obtained from both the Chin State government Department of Disaster Management , and Department of Meteorology and Hydrology to conduct a trial in Hakha. CCERR were responsible for engagement activities with the community for the duration of the trial. As a first step, they canvased the local community, including town elders, and confirmed a desire to participate in the project. Community representatives were thankful for the interest shown in helping them to address risks from landslides, which they live with every day. All parties demonstrated real interest, were happy to collaborate and provided enthusiastic support.

A key element of the trial design was to engage with the community, provide training on landslide causes and monitoring, and hence increase awareness and resilience to these catastrophic events. Core to this aim was a plan to engage with young people to act as Landslide Response Volunteers (LRVs), who would be trained to install and maintain CSS. This was accomplished by the Director of CCERR making a request for participants that was broadcast on local radio, which led to recruitment of 20 LRV in the age range 18 to mid-20 s, with 60% of LRV female. A member of CCERR staff acted as translator (between English and Burmese), enabling the LU team to explain activities on site and during training presentations, and production of translated CSS instructions and operation manuals.

Site selection

The 2015 Komen cyclone triggered large deep seated slope failures and many tens of relatively shallow small-scale landslides in the Hakha area of Chin State; both types caused destruction to property and infrastructure (Ministry of Mining and Myanmar Geo-Science Society 2015 ). When planning the project in 2017 there was limited information available on landslide hazards in the Hakha area to guide the selection of trial sites, other than the work of Thein ( 2016 ), which indicated that most of the Hakha area has a high to very high hazard designation. A study by Mon et al. ( 2018 ), which was published after the CSS trial commenced, details the geological setting and response to the landslide disaster in 2015.

The trial of CSS focussed on shallow failures that typically occur in the residual soils formed by weathering of the bedrock immediately beneath (mudstones and sandstones) and with a thickness in the order of 1 to 3 m. In places, these are overlain by thick layers of soil debris and colluvial from historical landslide activity. The dearth of local professionals who could advise on site selection meant the first step was to develop a selection procedure that employed local project representatives from CCERR and town elders, supported by FHI 360 project staff. The procedure was based on criteria established by the LU authors, who are experienced in studying landslide mechanisms. A site selection checklist (Table 1 ) was produced for use in Hakha by the local non-specialists. Based on seven criteria, it defined sets of questions written in non-technical language, most requiring either yes or no answers, that would provide the details required to rank and select sites with a risk of future shallow slope failure. Additional requirements and comments were provided to the assessors as context for the questions, and site photographs and GPS coordinates were also requested. This information was then used by LU to select the most promising sites for the installation and trial of CSS.

In November 2017, FHI 360 working with CCERR, state government and town elders held a workshop in Hakha to explain the planned trial and, based on collective knowledge and experience of the area, to produce a short-list of six candidate sites. These sites were then visited by the team of local representatives, and the checklist (Table 1 ) completed for each site and provided to LU. Feedback from the site review team indicated that the checklist was logical and easy to follow. LU selected two sites that best met the criteria, with two others designated as backups. Planning was progressed based on the two primary sites but with the proviso that their suitability would be reviewed and confirmed by LU as the first activity of the CSS installation campaign. The two selected sites were both in the Hakha urban area. The near surface geological materials at both sites comprise residual soils comprising fine sandy clayey SILT with some fine to coarse gravel.

Site 1 Keisi-Titawwin is in the middle of a slope that comprises an area of gardens, including a plant nursery, and scrub land hosting livestock (Fig.  4 ), bounded by an access track at one side, and houses and workshops at the other side and at the toe. The general undulating topography and cracking on the lower part of the site indicates previous shallow slope movements, which was confirmed by discussions with locals. The concern at this location is that during periods of heavy rainfall, movements could be reactivated, impacting on the properties at the toe and damaging the communal garden area.

Site 2 Keisi-Khaikam is at the top of a steep slope that has previously experienced landslides, as evidenced by the undulating slope profile and cracking. These failures damaged properties at the slope toe. The concern at this location is that retrogressive failures in the top of the slope would damage government buildings located close to the crest (Fig.  5 ).

General views of the slope at Keisi-Titawwin a the middle part of the slope and b the lower part of the slope

General view of the slope at Keisi-Khaikam, the buildings on the left would be impacted by any retrogressive slope failure in the slope to the right (i.e., the chain fence in the centre of the photo is located at the slope crest)

Both sites had easy and safe access, were covered in grass, small shrubs and occasional trees thus allowing good access and safe working conditions, permissions were in place from the local government, and they were located close to the homes of the LRV, thus giving short travel distances.

Description of monitoring system and installation process

The LU authors and staff from the FHI 360 Myanmar mission, visited Hakha in March 2018 to install the CSS systems and deliver training to CCERR and LRV members. Firstly, visits to the two selected sites (“ Site selection ” section) confirmed that both were suitable. Equipment and materials were in part delivered from the UK (e.g., CSS sensors, base station, and protective covers), with the remainder supplied locally based on specifications produced by LU (e.g., steel tubing for waveguides, post rammer, granular soil to infill the waveguides and laptop for system setup and data download).

The CSS systems were installed during a five-day period. The installation team comprised the LU authors, three project staff from FHI 360, two project officers of CCERR, the 20 youth LRV and ad hoc community engagement (e.g., a local carpenter made fences). During the works, many locals visited the sites to offer help and show interest. Local government representatives also visited the sites to observe the works and confirm their support. The LRV were trained by the LU authors to create and infill the waveguides, construct the cover systems, install sensors and base stations and then to set up, operate and maintain the monitoring system.

Figure  6 shows the system configurations and reference notation used at the two sites. Time and resources prohibited a detailed survey necessary to produce detailed plans or cross-sections. At Site 1, Keisi-Titawwin, Sensor 1 is in the upper part of the slope, and is closest to Base station 1, and Sensor 2 is in the lower part of the slope. Waveguides at each of the two locations were installed using the post rammer to drive the steel tubes to depths of 3 m below ground level. Sand infill was placed inside the tubes to form active ‘noisy’ waveguides (“ Community Slope SAFE (CSS) monitoring system ” section) and purpose-designed covers with pre-mounted solar panels were concreted in place to enclose and protect the waveguide and sensor. The sensors were attached to the waveguides using cable ties so that the piezoelectric elements were in good contact. The base station was housed in a lidded plastic container, with a solar panel positioned on top. All solar panels were orientated (i.e., rotated and tilted) to maximise solar gain. Figure  7 shows photos depicting the key stages of installation taken at the two sites.

CSS installation plans for Sites 1 and 2

Key stages of works by the authors, LRV and CCERR to install CSS at Sites 1 and 2: a driving the steel waveguide, b concreting the base for the cover, c placing the cover with help from interested children, d placing ‘noisy’ sand inside the steel waveguide, e the CSS sensor attached to the waveguide, f a completed sensor installation with integrated solar panel to charge the sensor battery, g set up of the base station and flashing light/siren alert system, h completed base station with solar panel to charge the battery

At Site 2, Keisi-Khaikam, both sensor nodes were installed behind the slope crest on horizontal ground, approximately 1.5 m from the top of the slope (Figs.  5 and 6 ). Sensor 3 is installed on a 5.5-m-deep waveguide and is closest to Base station 2, and Sensor 4 is installed on a waveguide 4-m-deep and is furthest away from the base station.

Before leaving Hakha, the LU authors commissioned the CSS system at each site and trained the CCERR liaison officer and LRV to operate the system and download data. Subsequently, the data was downloaded and emailed to LU by CCERR every two weeks for review. CCERR continued to act as a mentor to the LRV, providing support and training.

Operation and maintenance

Supported by LU, the LRV members and CCERR staff maintained and operated the CSS systems at the two sites from March 2018 until December 2019, when responsibility was passed to the Chin State government. However, the situation in Myanmar meant that the government department was not able to continue the monitoring. During that period, two-weekly site visits were made by CCERR and LRV to download data from each base station onto the laptop. In addition, operation of the system was checked by artificially generating AE on the waveguides to trigger a warning (i.e., by tapping a metal object on the top of the steel tube). The systems were re-initialised if communication between the sensors and base station had been lost (see below).

The data and reports sent by CCERR allowed LU to propose modifications to upgrade the system and improve robustness and performance for operation in the field environment. Upgrades planned by LU and implemented by CCERR/LRV in 2018/19 were carried out to:

Decrease background AE noise (i.e., interference)

Upgrade the wireless communication system, and

Improve battery charging.

Generally, the background levels of AE detected were well below the thresholds defined to trigger an alert of slope movement. However, there were periods when either general background levels increased or when peaks of AE were detected. After collecting data for a few months, analysis of the AE trends indicated that these increases were caused by electronic interference due to the waveguide acting as an antenna. Placing a thin strip of plastic tape between each piezoelectric transducer and steel tube isolated the sensor electronics from the waveguide and successfully eliminated this spurious AE.

The sensor nodes communicate with the base station at each site using a wireless system. This requires line of sight between the antenna at each sensor location and the base station. At both sites, adequate communication was checked before finalising the locations of the sensors and base station. However, an unforeseen government condition for permission to install CSS was that all elements (i.e., sensors and base station) must be protected by fences; once the wooden fences had been built around each installation, the wireless signal in all cases was reduced and communication was regularly lost. The wireless communication system was improved by CCERR/LRV installing new directional antennas (sent from the UK). In all cases, communication was improved and on the rare occasions it was lost, it could be re-established by artificially generating AE to trigger an alarm from the sensor to the base station. The upgraded antenna made CSS more reliable and resilient.

During prolonged periods of overcast weather in October/November 2018, the solar panels were unable to generate enough power to keep the sensors operational. These low battery levels were exacerbated by the communication issues detailed above, which used additional power as the sensors tried repeatedly to re-connect to the base station. This issue was overcome by upgrading the wireless antennas to reduce power consumption and repositioning the solar panels to optimise efficiency. No further losses of power occurred in 2019 after these modifications. Work by CCERR and the LRV to clear vegetation every few months from around the installations also aided performance of the system. Dense vegetation can shade the solar panels and interfere with wireless signals.

In 2019, following completion of the above system upgrades, the CSS monitoring systems at the two sites operated as designed, with only small numbers of sensor/base station communication issues experienced.

Example measured AE data is presented in Fig.  8 a) for Site 2, Sensor 3. It shows very low levels of AE detected throughout the monitoring period. This is indicative of stable ground. The peaks in AE at the start and end were artificially generated by tests conducted during LRV visits. Because no slope instability events occurred at the sites during the trial, there was not an opportunity to detect ground movements using CSS and hence test triggering of the visual/audible alert system. However, the trial was still able to provide useful learning as discussed in “ CSS performance ” section. Figure  8 b) presents example data for operation of the solar panel (voltage level) and charging of the battery (percent). This shows the system operating continuously over a period of 5 months.

Site 2, Sensor 3 a Example measured AE data showing low level aggregated AE RMS time series measurements indicative of stable ground. The peaks of AE at the start and end of the monitoring period were generated by CCERR/LRV visiting the site to trigger an alert and test operation of the system. b shows example solar (V) and battery level (%) data. (Dixon and Smith 2022 )

The AE warning threshold level shown in Fig.  8 (i.e., the horizontal dashed line) was set such that exceedance would trigger an alert of potential movements. Based on experience from monitoring comparable slopes, an AE threshold was selected to warn of displacement rates in the order of 5 mm per minute. Laboratory calibration for the CSS system by Dixon et al. ( 2018 ) was used to define the AE RMS level indicative of this magnitude of displacement rate. Despite no slope failures occurring, the 20 + months of monitoring experience allowed the selected AE trigger level to be reviewed. It was not exceeded during normal operation and no false alerts were generated.

Community training and engagement

An important benefit of the CSS trial in Hakha has been raising awareness in the community to improve resilience to future landslide events. CCERR led this work, aided by the LRVs, by generating publicity about CSS, running a series of education events, and holding regular liaison meetings with the partner Chin State government Departments of Disaster Management, and Meteorology and Hydrology . Signs were erected at the CSS monitoring sites to highlight the project and to publicize the involvement of the local organizations (Fig.  9 ). Media organizations reported on the project.

CCERR with LRV and FHI 360 author publicising the landslide monitoring project via local media coverage

Six open community training events were run in 2018 by CCERR and the LRV who were tasked with championing landslide awareness in the community. The aim was to engage with and educate the community from Hakha township and the municipal ward. About 80 people from Hakha and surrounding communities participated in these events to learn about the CSS sensor programme, discuss landslide risks, consider appropriate community response, and hence improve resilience to landslides. Through these events, CCERR (2018) reported that participants were tasked to ‘ disseminate about sensors further to the wider community …. during worship service, at churches or houses and other social occasions such as weddings and funerals… ’. These community engagement activities created a large group of trained landslide ambassadors in Hakha and the surrounding region, changing behaviour and delivering improved community awareness and resilience to future landslide events.

The local partners understood that this trial was part of a longer process to learn about the CSS approach, improve sensor performance and hence enable future benefits for the wider Myanmar community. In recognition of the training received and contribution to landslide monitoring and awareness in Hakha, the Landslide Response Volunteers were awarded certificates, which acknowledge and celebrate their valuable contribution to the project under the commendation ‘ Myanmar Youth use Sensors to Save Lives ’. In a follow up visit by FHI 360 in summer 2018, feedback from CCERR and LRV was that their involvement in the trial was a very positive experience, and they were grateful for the opportunity and proud to play a significant role in what they consider to be an important initiative.

In a follow-up review of the project impacts by LU, CCERR (2020) reported that the project not only allowed the community to overcome a collective psychological panic over landslides, but it also demonstrated the government could take scientific measures to address landslide risks. Community leaders such as the Hakha Town Elders, a committee for Hakha Town's Affairs, came to appreciate and take ownership of the sensors. It is estimated that around 15,000 to 20,000 population in Hakha town benefited from the trial (CCERR 2020).

However, because monitoring was discontinued by the state government department, the system has not reached full operational status (i.e., with the alert system live), and therefore a local action plan is yet to be established detailing how the community should respond to an alert (e.g., communication and evacuation plan). Despite this, the evidence is that landslide awareness and knowledge in the community was increased by the trial. It has also been demonstrated that non-specialist local groups (e.g., CCERR and LRV) can be trained to install and operate the CSSsystem and to act as landslide ambassadors within the community.

This has raised interest and provided a positive example of how it is possible to undertake programs that deliver benefits via scientific (e.g., LU) and development (e.g., FHI 360 and CCERR) sectors working together to improve the lives of communities in low-income countries. A unique impact of this collaboration was being able to align engagement of the central government of Myanmar all the way through the state government and CCERR agency to the youth volunteers and ultimately the community. This project called national government attention to the fact that there are ways the local communities can help make significant positive impacts on a major issue such as reducing landslide risk.

CSS performance

The trial of CSS slope monitoring in a remote part of the world using inexperienced local groups to install and then operate the system for 20 + months was a significant achievement by the partners. In addition to the community benefits detailed in “ Community training and engagement ” section, there have also been useful learning on a range of practical aspects associated with planning, installing, and operating the system. It is important to recognise the very different physical operational conditions between the UK, where CSS was devised and initially tested, and Hakha (e.g., higher maximum temperatures, different solar radiation patterns and faster vegetation growth rates). Key learning from the trial includes:

The imperative to ensure the community has access to relevant information on landslide risks so that priority sites for monitoring can be identified. In this trial, availability of limited hazard mapping for Hakha was supplemented by input from the LU landslide experts in the form of a pro-forma check list for site assessment.

Gaining the approval and conditions of all parties to undertake the work is obvious, but a complex task that will vary according to the local social norms and governance. Slope monitoring cannot be achieved without engagement of local organisations from the initial proposal stage. It cannot be ‘imposed’ by an external party, even if they have the best intentions. For this trial, it took FHI 360 and CCERR many months to gain permissions from the national and regional governments.

Location of sensors on a given site requires considerable experience and knowledge of landslide processes. Local groups will require this level of landslide expertise/support to establish slope monitoring projects, even if they have the knowledge to install and operate a system. Local and national government agencies should have the expertise available and processes in place to support the community.

Local people were able to quickly learn the skills required to install the waveguides and covers. Driving the steel tube waveguides to depth was physically demanding using post rammers and it is recommended that manually operated tripod drop weight systems are employed in the future. Account must also be made for local norms and practices when considering what is acceptable health and safety practices.

The knowledge and skills to operate the CSS system (i.e., set up, download data and trouble shoot problems) was achieved by the local partners through the formal training sessions coupled with on-site instruction provided by LU. Given that the LU authors were only on site for a week, it is encouraging that CCERR with LRV were able to operate the system for 20 + months, including helping to diagnose faults and install upgrade components.

Following completion of the upgrades detailed in “ Operation and maintenance ” section, the system operated as designed:

AE was measured continuously by the sensors

Low levels of AE were synonymous with stable slope conditions

Measured AE was automatically compared to predefined trigger thresholds

Experiments to artificially generate high levels of AE that exceeded the trigger threshold resulted in automatic communication between sensor and base station to generate an alert

No false alarms were generated during normal operation, demonstrating robustness

Solar panels were able to charge the sensor and base station batteries to achieve continuous operation

Daily health messages (e.g., battery charge level) were generated automatically by the sensors and communicated to the base station where they were logged.

No cases of damage or interference with the sensors were reported, which is a testament to the community engagement and collective guardianship of the monitoring system.

The project demonstrated that installation and operation of a slope EWS such as CSS is both achievable and beneficial for the education, protection and wellbeing of a community.

Conclusions

A low-cost landslide early warning system that monitors acoustic emission (AE) generated by subsurface slope movements has been developed, called Community Slope SAFE. This paper described the CSS approach and detailed a field trial of the system at two sites in Hakha, Chin State, Myanmar. The trial was delivered via a partnership between the developers of CSS at Loughborough University (LU) who are landslide experts, FHI 360 an international organization working to improve the health and well-being of people in Myanmar, the Chin Committee for Emergency Response and Rehabilitation (CCERR) and both state and national government departments in Myanmar. The following are the principal conclusions:

A site selection check list produced by LU enabled local non landslide specialists to identify candidate sites, which led to the successful selection of two sites for installation of the CSS system.

During a one-week period in March 2018, the LU team were able to train representatives from CCERR and a group of 20 youth Landslide Response Volunteers (LRVs) to install and operate the CSS systems.

CCERR and LRV successfully operated the monitoring system, including sending data to LU for review, from March 2018 until December 2019.

The specific CSS operating conditions in Hakha meant that modifications to key elements of the system were required. These included upgrading the wireless communication and optimising the solar panel charging of batteries. Key learning from this process is that local non-specialist CCERR and LRV personnel were capable of undertaking modifications and maintenance to the system, guided via instructions from LU.

Performance of the monitoring system was demonstrated by its operation for months as designed and without problems. There were no slope movements at the monitoring sites during the trial and therefore the system did not experience full operational conditions. However, consistently low levels of detected AE were recorded, indicative of stable slopes, and no false alarms were triggered.

A key benefit, which is independent of the type of monitoring system employed, has been using this demonstration project to engage with the Hakha community and significantly enhance awareness of landslide risks.

Training provided by LU to CCERR and the LVR created a group of landslide champions who engaged with multiple community groups and energised awareness and discussion of landslide risks and monitoring.

AE sensors for monitoring slopes are now commercially available and therefore an opportunity exists to replicate the community slope monitoring approach globally. However, this will require sponsors to fund purchase of the monitoring instruments and provide expertise to select sites, and to train communities to install and operate the system.

The project partnership delivered the Community Slope SAFE trial and demonstrated technical advances in slope monitoring for communities, but potentially more important, the trial has delivered positive impact on lives in the Hakha community and shown that this could be replicated in other communities.

Availability of data and materials

The datasets presented in Fig.  8 can be found at https://doi.org/10.17028/rd.lboro.19382939 , an open-source online data repository hosted at Loughborough University (Dixon and Smith 2022 ).

Berg B, Smith S, Russell S, Dixon N, Proudfoot D, Take WA (2017) Correlation of acoustic emissions with patterns of movement in an extremely slow-moving landslide at Peace river, Alberta. Canada Canadian Geotech J 55(10):1475–1488. https://doi.org/10.1139/cgj-2016-0668

Article   Google Scholar  

Chichibu A, Jo K, Nakamura M, Goto T, Kamata M (1989) Acoustic emission characteristics of unstable slopes. J Acoustic Emission 8(4):107–112

Google Scholar  

de Assis Dias MC, Saito SM, dos Santos Alvalá RC, Seluchi ME, Bernardes T, Camarinha PIM, Nobre CA (2020) Vulnerability index related to populations at-risk for landslides in the Brazilian early warning system (BEWS). Int J Disaster Risk Reduction 49:101742. https://doi.org/10.1016/j.ijdrr.2020.101742

Dixon N, Spriggs MP, Smith A, Meldrum P, Haslam E (2015a) Quantification of reactivated landslide behaviour using acoustic emission monitoring. Landslides 12(3):549–560. https://doi.org/10.1007/s10346-014-0491-z

Dixon N, Smith A, Spriggs MP, Ridley A, Meldrum P, Haslam E (2015b) Stability monitoring of a rail slope using acoustic emission. Proc Inst Civ Eng Geotech Eng 168(5):373–384. https://doi.org/10.1680/jgeen.14.00152

Dixon N, Smith A, Flint JA, Khanna R, Clark B, Andjelkovic M (2018) An acoustic emission landslide early warning system for communities in low-income and middle-income countries. Landslides 15(8):1631–1644. https://doi.org/10.1007/s10346-018-0977-1

Froude MJ, Petley DN (2018) Global fatal landslide occurrence from 2004 to 2016. Nat Hazard 18(8):2161–2181. https://doi.org/10.5194/nhess-18-2161-2018

Kalubowila P, Lokupitiya E, Halwatura D, Jayathissa G (2021) Threshold rainfall ranges for landslide occurrence in Matara district of Sri Lanka and findings on community emergency preparedness. Int J Disaster Risk Reduction 52:101944. https://doi.org/10.1016/j.ijdrr.2020.101944

Koerner RM, McCabe WM, Lord AE (1981) Acoustic emission behaviour and monitoring of soils. In Acoustic Emission in Geotechnical Practice, ASTM STP 750:93–141

Liu Y, Yin K, Chen L, Wang W, Liu Y (2016) A community-based disaster risk reduction system in Wanzhou, China. Int J Disaster Risk Reduction 19:379–389. https://doi.org/10.1016/j.ijdrr.2016.09.009

Michlmayr G, Chalari A, Clarke A, Or D (2017) Fiber-optic high-resolution acoustic emission (AE) monitoring of slope failure. Landslides 14(3):1139–1146. https://doi.org/10.1007/s10346-016-0776-5

Mon MM, Naing T, Numada M, Yu KT, Meguro K, Latt KZ (2018) Analysis of disaster response during landslide disaster in Hakha, Chin State of Myanmar. J Disaster Res 13(1):99–115. https://doi.org/10.20965/jdr.2018.p0099

Nam K, Wang F (2020) An extreme rainfall-induced landslide susceptibility assessment using autoencoder combined with random forest in Shimane Prefecture. Japan Geoenvironmental Disasters 7(1):1–16

Ngandam Mfondoum AH, Wokwenmendam Nguet P, Mefire Mfondoum JV, Tchindjang M, Hakdaoui S, Cooper R, Moudioh C (2021) Adapting sudden landslide identification product (SLIP) and detecting real-time increased precipitation (DRIP) algorithms to map rainfall-triggered landslides in Western Cameroon highlands (Central-Africa). Geoenvironmental Disasters 8(1):1–26

Smith A, Dixon N (2015) Quantification of landslide velocity from active waveguide-generated acoustic emission. Can Geotech J 52(4):413–425. https://doi.org/10.1139/cgj-2014-0226

Smith A, Dixon N, Meldrum P, Haslam E, Chambers J (2014) Acoustic emission monitoring of a soil slope: Comparisons with continuous deformation measurements. Géotechnique Letters 4(4):255–261. https://doi.org/10.1680/geolett.14.00053

Smith A, Dixon N, Fowmes GJ (2017a) Early detection of first-time slope failures using acoustic emission measurements: large-scale physical modelling. Géotechnique 67(2):138–152. https://doi.org/10.1680/jgeot.15.P.200

Smith A, Dixon N, Moore R, Meldrum P (2017b) Photographic feature: acoustic emission monitoring of coastal slopes in NE England, UK. Q J Eng GeolHydrogeol 50(3):239–244. https://doi.org/10.1144/qjegh2016-081

Stähli M, Sättele M, Huggel C, McArdell BW, Lehmann P, Van Herwijnen A, Berne A, Schleiss M, Ferrari A, Kos A, Or D (2015) Monitoring and prediction in early warning systems for rapid mass movements. Nat Hazard 15(4):905–917. https://doi.org/10.5194/nhess-15-905-2015

Chin committee for emergency response and rehabilitation (2018) Report for FHI 360, October (unpublished)

Chin committee for emergency response and rehabilitation (2020) Impact of landslide monitoring project, 5th October. Personal communication from the Director.

Cuai HT (2017) lessons learned from landslide disaster recovery in Hakha Town, and how to strengthen resilience. PPT presentation (Unpublished) https://static1.squarespace.com/static/575fb39762cd94c2d69dc556/t/596a3867a803bb6c0a34fffb/1500133498161/Hlawn+Tin+Cuai.pdf (accessed 27/1/2022)

Dixon N, Smith A (2022) Acoustic emission RMS, solar voltage and battery charge percentage time series data for Site 2 Sensor 3 of the Community Slope SAFE trial at Hakha, Chin State, Myanmar. Loughborough Data Repository, https://doi.org/10.17028/rd.lboro.19382939

Fujiwara T, Ishibashi A, Monma K (1999) Application of acoustic emission method to Shirasu slope monitoring. In Slope stability engineering (Yagi, Yamagami & Jiang (eds)). Rotterdam: Balkema, pp. 147–150.

Kjekstad O, Highland L (2009) Economic and social impacts of landslides. In: K Sassa, P Canuti (eds.), Landslides – disaster risk reduction, Springer-Verlag Berlin Heidelberg.

Ministry of Mining and myanmar geo-science society (2015) Engineering geological report of landslide and road subsidence at Hakha City and Its Environs.

Nakajima I, Negishi M, Ujihira M, Tanabe T (1991) Application of the acoustic emission monitoring rod to landslide measurement. Proc. 5th Conference on acoustic emission/microseismic activity in geologic structures and materials, Pennsylvania, pp. 1–15.

Thein KSM (2016) Rain-triggered landslide hazard mapping in soft rock mountainous region: a case study in Chin State, Myanmar. Doctoral candidate, asian institute of technology, Thailand, (Unpublished)

UNEP (2012) Early warning systems: a state-of-the-art analysis and future directions. division of early warning and assessment (DEWA), United Nations environment programme (UNEP), Nairobi, 2012.

UNISDR (2007) Hyogo framework for action (HFA) 2005–2015: building the resilience of nations and communities to disasters. United Nations international strategy for disaster risk reduction (UNISDR), Geneva, Switzerland, 28pp.

UNISDR (2009) Terminology on disaster risk reduction, united nation international strategy for disaster reduction. united nations international strategy for disaster reduction (UNISDR), Geneva, Switzerland, 35 pp.

UNISDR (2015) Sendai framework for disaster risk reduction 2015 – 2030. united nations international strategy for disaster reduction (UNISDR), UNISDR/GE/2015 - ICLUX EN5000, 1st edition.

Download references

Acknowledgements

Special thanks are due to the staff of FHI 360—Ye Htut, Program Director, Yu War and Daung Zay, Project Operatives, for planning and coordinating the field work, and to CCERR Director Flora BN Mawi and Liaison Officers Philip Nun Uk Thang and Lian Cung Mang for delivering the community engagement activities, recruiting the Landslide Response Volunteers, organising training events and overseeing operation of CSS. Thanks are also due to Datalink Electronics Ltd. for developing the CSS system electronics.

Development of the Community Slope SAFE monitoring system was funded by UK Engineering and Physical Sciences Research Council, Loughborough University Impact Acceleration Awards. The field trial in Hakha, including further sensor development, was funded by the FHI 360 FY16 Catalyst Fund. Alister Smith acknowledges the support of a Philip Leverhulme Prize (PLP-2019–017).

Author information

Authors and affiliations.

School of Architecture, Building and Civil Engineering, Loughborough University, Leicestershire, LE11 3TU, UK

Neil Dixon & Alister Smith

Civil Society & Media Project II in Myanmar, FHI 360, 359 Blackwell St #200, Durham, NC, 27701, United States of America

Matthew Pietz

You can also search for this author in PubMed   Google Scholar

Contributions

ND and AS conceived the CSS system and carried out the laboratory and field work, and analysis and interpretation of the data. MP coordinated the field trial locally from Myanmar. All authors secured funding. All authors read and approved the final manuscript.

Corresponding author

Correspondence to Alister Smith .

Ethics declarations

Competing interests.

The authors declare that they have no competing interests.

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Dixon, N., Smith, A. & Pietz, M. A community-operated landslide early warning approach: Myanmar case study. Geoenviron Disasters 9 , 18 (2022). https://doi.org/10.1186/s40677-022-00220-7

Download citation

Received : 06 June 2022

Accepted : 10 August 2022

Published : 24 August 2022

DOI : https://doi.org/10.1186/s40677-022-00220-7

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Acoustic emission
• Community preparedness/Resilience
• Early warning

Quick links

• Climate change
• Multiple hazards
• Decarbonisation

Folkestone Warren, Kent

Landslide case study

The Folkestone Warren landslide is a very large, deep-seated coastal landslide that is well known to geologists and engineers. It is about 3 km wide and up to 350 m in length. There are nine surveys in the BGS National Landslide Database : IDs 1774/1 to 1774/9.

The village of Capel-le-Ferne is situated behind the landslide at the cliff top and the wide undercliff is given over to a country park.

Folkestone Warren location map. BGS © UKRI.

The Dover to London mainline railway (via Cheriton) passes through the entire length of the Warren, with the Abbotscliff Tunnel at the eastern end and the Martello Tunnel at the western end. Since its construction, the railway has been regularly disrupted by landslide activity and considerable expenditure has been, and continues to be, involved in remediation and monitoring. The old Dover road is situated close to the cliff edge. This has also been disrupted, and re-routed in parts, for the same reason.

The geology of the platform at beach level varies along the length of the Warren. At the eastern end, the platform changes from Gault Formation to chalk. The change occurs approximately in line with the eastern end of the concrete toe-weight apron (constructed in 1953), near Horsehead Point. Rock revetment has been placed to the east of the concrete apron by Network Rail. It is notable that, westward of Horsehead Point, the platform is displaced seaward by about 50 m due to movements of the various slide masses (Trenter and Warren, 1996).

Figure 1 The Folkestone Warren Landslide (Photograph taken September 2002)

The geology consists of the White and Grey Chalk subgroups overlying the Gault Formation and the Folkestone Formation , the principal slide surface being located at the Gault/Folkestone junction.

The upper part of the chalk sequence here is the White Chalk Subgroup comprising, from top to bottom, the New Pit Chalk Formation resting on the Holywell Nodular Chalk Formation .

The lower part of the chalk sequence is the Grey Chalk Subgroup , formed by the Zig Zag Chalk Formation resting on the West Melbury Marly Chalk Formation . Superficial deposits forming the Clay-with-Flints Formation overlie the chalk.

Beneath the chalk sequence lies the Gault Formation, with the thin ‘Chloritic Marl’ bed at their junction. The Gault Formation consists of between 44 and 48 m of overconsolidated clay. Beneath the Gault Formation is the Folkestone Formation of the Lower Greensand Group , with the ‘Sulphur Band’ at their junction. The Folkestone Formation consists of calcareous and glauconitic sands and sandstones with some argillaceous horizons (Trenter and Warren, 1996).

The Folkestone Warren landslide is one of the largest on the English coast and is a classical example of a deep-seated (approximately 40 m below sea level) multiple retrogressive, compound mechanism, having translational , rotational and graben (sunken block) features.

The Folkestone Warren cliffs are divided into the chalk ‘High Cliff’, essentially the overall landslide backscarp, and the sea ‘cliff’, consisting largely of landslipped material. The High Cliff as such is not directly influenced by the sea, unlike the Folkestone Warren landslide as a whole.

The landslide has been extensively studied since 1915, when a major re-activation occurred throughout the complex seriously disrupting the railway that was constructed in 1844 (Hutchinson, 1969; Hutchinson et al., 1980; Trenter and Warren, 1996; Warren and Palmer, 2000). Included in this event were chalk slides from the High Cliff, in particular the ‘Great Fall’ landslide. Hutchinson et al. (1980) describes the three dominant subvertical joint sets within the chalk and their influence on the form of the High Cliff and on landslides within it. In recent decade,s a large amount of instrumentation and remedial works has been installed and maintained by Network Rail within the Warren.

Massive rotational failure in 1915 affecting the Folkestone to Dover train (British Railways, Southern Region).

The age of the original landslide event is not known. Most landslide movements in recent times have been confined to the undercliff in the form of re-activations of existing, deep-seated landslide masses. However, numerous significant slope failures of the High Cliff are believed to have occurred since the 1800s.

As part of a programme of work monitoring coastal erosion and landsliding at several sites around the coast of Great Britain, BGS has surveyed the Folkestone Warren landslide complex with LiDAR .

The Folkestone Warren Landslide. BGS © UKRI (taken September 2002).

Folkestone Warren cliffs. BGS © UKRI.

Folkestone Warren from railway bridge. BGS © UKRI.

Folkestone Warren landslide (east). BGS © UKRI.

More information

Hutchinson, J N. 1969. A reconsideration of the coastal landslides at Folkestone Warren, Kent. Géotechnique , Vol. 19(1), 6–38.

Hutchinson, J N, Bromhead, E N, and Lupini J F. 1980. Additional observations on the Folkestone Warren landslides . Quarterly Journal of Engineering Geology & Hydrogeology , Vol. 13(1), 1–31. DOI: https://doi.org/10.1144/GSL.QJEG.1980.013.01.01

Trenter, N A, and Warren, C D. 1996. Further investigations at the Folkestone Warren landslide. Géotechnique , Vol. 46(4), 589–620.

Warren, C D, and Palmer, M J. 2000. Observations on the nature of landslipped strata, Folkestone Warren, United Kingdom. Proceedings of the 8th International Symposium on Landslide s, Cardiff, June 2000.

Acland, C L. 1869. The late landslip on the Warren. Quarterly Journal of the Folkestone Natural History Society , Vol. 3, 48–51.

Arber, M A. 1940. The coastal landslips of south-east Devon . Proceedings of the Geologists’ Association , Vol. 51, 257–271.

Barton, M E, and McCosker, A M. 2000. Inclinometer and tiltmeter monitoring of a high chalk cliff. 127–132 in Landslides in Research, Theory and Practice . (Bromhead, E, Dixon, N, and Ibsen, M-L (editors). (London, UK: Thomas Telford.)

Bell, F G, Cripps, J C, Edmonds, C N, and Culshaw, M G. 1990. Chalk fabric and its relation to certain geotechnical properties. 187–193 in Chalk .

Bishop, C H. 1973. Folkestone: The Story of a Town. (London and Ashford, UK: Headley Brothers Ltd.)

Bromhead, E N, Hopperm A C, and Ibsen, M-L. 1998. Landslides in the Lower Greensand escarpment in south Kent . Bulletin of Engineering Geology and the Environment , Vol. 57(2), 131–144.

Bromhead, E N, and Ibsen, M-L. 2004. Bedding-controlled coastal landslides in Southeast Britain between Axmouth and the Thames Estuary . Landslides , Vol. 1, 131–141.

Drew, F. 1864. The geology of the country between Folkestone and Rye, including the whole of Romney Marsh. Memoir of the Geological Survey of Great Britain. (London, UK: HMSO.)

Folkestone Chronicle, 10 December, 1859.

Folkestone Chronicle, 11 February 1856.

Folkestone Chronicle, 20 January 1877.

Gentleman’s Magazine. 1806. Country News (June). Vol. 76, 575.

Haselock, P J, and Gale, A R. 1998. The Folkestone Warren landslip: a model for the decay of fault scarps? Proceedings of the 19th Annual Meeting of the Tectonic Studies Group .

Hutchinson, J N. 1965. A survey of coastal landslides in Kent. Building Research Station Note No 35/36.

Hutchinson, J N. 1986. Cliffs and shores in cohesive materials; geotechnical and engineering geological aspects. 1–44 in Proceedings of the Symposium on Cohesive Shores , 5–7 May 1986 Burlington, Ontario. Skafel, M G (editor).

May, V J. 1964. A study of recent coastal changes in southern England. MSc thesis (unpublished) University of Southampton.

McDakin, J G. 1900. Coast Erosion: Dover Cliffs.

McDakin, J G. 1911. Abbotscliff (Warren) Chalkfall Geological Notes. East Kent Scientific and Natural History Society Reports and Transactions , Vol. 16, 15.

McGowan, A, Roberts, A G, and Woodrow, L K R. 1988. Geotechnical and planning aspects of coastal landslides in the United Kingdom. In: Landslides Glissements de Terrain . Bonnard, C (editor). Vol. 2.

Morgenstern, N R, and Price, V E. 1965. The analysis of the stability of general slip surfaces. Géotechnique , Vol. 15, 79–93.

Muirwood, A M. 1971. Engineering aspects of coastal landslides. Proceedings of the Institution of Civil Engineers , Vol. 50, 257–276.

Osman, C W. 1917. The landslips of Folkestone Warren and thickness of the Lower Chalk and Gault near Dover. Proceedings of the Geologists’ Association , Vol. 28, 59–84.

Phillips, W. 1821. Remarks on the Chalk cliffs in the neighbourhood of Dover, and on the Blue Marls covering Green Sand near Folkestone. Transactions of the Geological Society , Vol. 5, 16–46.

Price, F G H. 1877. On the beds between the Gault and Upper Chalk near Folkestone. Quarterly Journal of the Geological Society , Vol. 33, 431–448.

The Railway Magazine. 1919. Railway landslides: engineering problems at Wembley and the Warren. July issue, 45: 23–31.

The Times, 14 December 1859, page 12.

Toms, A H. 1946. Folkestone Warren landslips: research carried out in 1939 by the Southern Railway Company. Institute of Civil Engineers , Railway Paper No. 19.

Topley, W. 1893. The landslip at Sandgate. Proceedings of the Geologists’ Association , Vol. 13(2), 40–47.

Viner-Brady, N E V. 1955. Folkestone Warren landslips: remedial measures 1948–1954 . Proceedings of the Institution of Civil Engineers, Vol. 4(3), 429–441.

Wood, M. M. (1955), Folkestone Warren landslips: investigations 1948–1950 . Proceedings of the Institution of Civil Engineers , Vol. 4(3), 410–428.

Contact the Landslide Response Team

You may also be interested in.

Landslide case studies

The landslides team at the BGS has studied numerous landslides. This work informs our geological maps, memoirs and sheet explanations and provides data for our National Landslide Database, which underpins much of our research.

Understanding landslides

What is a landslide? Why do landslides happen? How to classify a landslide. Landslides in the UK and around the world.

How to classify a landslide

Landslides are classified by their type of movement. The four main types of movement are falls, topples, slides and flows.

Landslides in the UK and around the world

Landslides in the UK, around the world and under the sea.

Was this page helpful?

How can we make this section better?*

• How can we make this section better? *

Please select a reason*

• Please select a reason * Select reason Didn't answer my question Too confusing Links are broken/Doesn't work

Academia.edu no longer supports Internet Explorer.

To browse Academia.edu and the wider internet faster and more securely, please take a few seconds to  upgrade your browser .

Enter the email address you signed up with and we'll email you a reset link.

• We're Hiring!
• Help Center

A CASE STUDY ON A LANDSLIDE: A GEOTECHNICAL INVESTIGATION

https://irjet.net/archives/V3/i4/IRJET-V3I4595.pdf

Related Papers

Miroslav Todorov , Mihail Todorov

Landslide processes are a distinct phenomenon with a huge impact on human way of life. Those gravitational movements in the upper layers of the crust can have a dramatic effect in river valleys, coastal regions and mountainous terrains. The reasons for their existence and development lie in naturally occurring processes in the lithosphere and the atmosphere and large-scale human activity. Investigating various natural and anthropogenic phenomena is vital, with attention to climate change processes. In the report of the European Commission (2012) an alarming level of erosion and abrasion is observed, with more than 630 000 areas with landslide processes. According to MRRB in Bulgaria there are near 2000 zones with active movements. The current investigation focuses on a fragment of the Bulgarian coast prone to landslide that was researched in 2017. The first registered movements date back to 2010. Characterizing for this case is the relatively quick developments of the movements which affect urban areas. This publication revolves around a complex engineering investigation, which provides an understanding how and why activation processes in the researched zone form, while using well-known proven methods. The used techniques for investigation focus on usability issues, as well as on the opportunity to corelate different sets of data, gather important information and reinforce it with monitoring. The focus here is on these three combined steps: information, prediction, monitoring.

Peter Bobrowsky

Regardless of the exact definition used or the type of landslide under discussion, understanding the basic mechanics and processes of a typical landslide is critical. With an increasing awareness and mandate to fully understand new building site characteristics, improve existing critical infrastructure and appreciating the importance of an area’s landslide history, typing landslides according to shared characteristics gives vital information about the future performance of a site, in relation to potential landslide hazards. We provide here, a basic primer for an understanding of the similarities and differences of the nature and physics of landslide movement. Type of material, speed of motion, slope angle, potential frequency of occurrence, weather and climatic influences, and man-made disturbances as well as other factors have a bearing on landslide motion, size and impact, yet we can generally group and categorize most landslides into more understandable groupings. Landslide typol...

Advancing Culture of Living with Landslides

Željko Arbanas

International Journal of Geophysics

Stefano Morelli , Veronica Pazzi

Geohazards processes can damage or increase the risk of human beings, properties, critical infrastructures, and environment itself. ey also could involve the interruption of human activities with serious socioeconomic consequences. Among all the natural occurrences, landslides are regarded as one of the most destructive types of geohazards. Landslides are a type of "mass wasting," which denotes any down-slope movement of soil and rock under the direct influence of gravity, which can occur and develop in a large variety of volumes and shapes. Even though the catastrophic impact of landslides is not totally unavoidable, it can be significantly reduced by increasing the capacity to assess and predict the risks and using different mitigation methods. In the past decades, many D and D numerical modelling methods have been designed and developed to assess slope stability, to predict slope response to various triggers, to evaluate the slope deformation and evolution pattern, and to perform back-analysis simulations. Nevertheless, such models still require access to detailed knowledge of the geological, mechanical, and hydrological properties of landslides and boundary conditions. erefore, accurate geological field surveys have to be integrated by means of low-cost and noninvasive techniques, like the geophysical ones, to collect widespread data with the aim of reconstructing a suitable geological and hydrogeological model of the area, improving the reliability of deterministic model. is special issue is dedicated to the geophysical methods applied to investigate, characterize, and monitor landslides. Over the years, both the advantages and limitations of these techniques have been highlighted, and some drawbacks are still open. Some papers were submitted to this special issue, and, aer a thorough peer review process, only five articles were selected to be included in this special issue. is relatively small number is probably caused by the difficulty in applying geophysical techniques on slope movements given hard-operating conditions (e.g., high slopes, distance from access roads, and lack of security for the technical operator) and not because the methods limitations are greater than the advantages. e review carried out by V. Pazzi et al. on geophysical techniques applied in landslides studies analyses the international efforts toward overcoming the geophysical technique limitations highlighted by the geophysics and landslide review, focusing on works of the last twelve years (-). e authors carried out the review analysis using a "material landslide approach" on the basis of the more recent landslides classification. e most studied landslides are those of the flow type for "soil" landslide typology and those of the fall type for the "rock" category. From the "employed method" point of view, active and passive seismic methods are the most employed in landslide characterization and monitoring. To quantify the efforts performed to overcome the limitations highlighted in , a three-level scale was employed (from many/some efforts to non-discussed). e limits inherent in each technique and the need to still develop multisource data integration methods were clear; very oen

Fabio Vittorio De Blasio

stefano utili

Landslide Science for a Safer Geoenvironment

Jose Chacon

RELATED PAPERS

Acta Polymerica

Yitzhak Rabin

Antonio Di Natale

Infection and Drug Resistance

Bert Gerrits van den Ende

Research in Science Education

Miloud Benhamou

Bulletin des Sociétés Chimiques Belges

Nitin W Fadnavis

European heart journal

Andreas Eicken

Journal of Lipid Research

Uporabna informatika

Katarina Puc

Revista De La Facultad De Ciencias Agrarias

Patricia Hashimoto

Priscila Tavares semprebom

Ciência Rural

WALTER PEREIRA

The Journal of Physical Chemistry C

Yaroslav Filinchuk

Jack Stetter

Lic. Julio Zetter Leal

2013 IEEE International Conference on Acoustics, Speech and Signal Processing

Charles Taylor

Fithnatia Khoirunnisa

Narada : Jurnal Desain dan Seni

Lingga Partawijaya

2015 Sixth International Conference on Emerging Security Technologies (EST)

Gareth Howells

Luis Torres

Bulletin of the American Physical Society

Nazarena Spera

Jose Roberto Cardoso

Journal of Air Law and Commerce

Griselda Capaldo

Ahi Evran Üniversitesi Sosyal Bilimler Enstitüsü Dergisi Cilt 1, Sayı 1, 2014

Mahmut Bolat

RELATED TOPICS

•   We're Hiring!
•   Help Center
• Find new research papers in:
• Health Sciences
• Earth Sciences
• Cognitive Science
• Mathematics
• Computer Science
• Academia ©2024

• Reference Manager
• Simple TEXT file

People also looked at

Original research article, the landslide traces inventory in the transition zone between the qinghai-tibet plateau and the loess plateau: a case study of jianzha county, china.

• 1 National Institute of Natural Hazards, Ministry of Emergency Management of China, Beijing, China
• 2 Key Laboratory of Compound and Chained Natural Hazards Dynamics, Ministry of Emergency Management of China, Beijing, China
• 3 Key Laboratory of Shale Gas and Geoengineering, Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing, China
• 4 Institute of Geology, China Earthquake Administration, Beijing, China

The upper reaches of the Yellow River in China, influenced by erosion of the Yellow River and tectonic activities, are prone to landslides. Therefore, it is necessary to investigate the existing landslide traces. Based on visual interpretation on high-resolution satellite images and terrain data, supplemented and validated by existing landslide records, this paper prepared the most complete and detailed landslide traces inventory in Jianzha County, Huangnan Tibetan Autonomous Prefecture, Qinghai Province, to date. The results indicate that within the study area of 1714 km 2 , there are at least 713 landslide traces, ranging in scale from 3,556 m 2 to 11.13 km 2 , with a total area of 134.46 km 2 . The total landslide area excluding the overlap area is 126.30 km 2 . The overall landslide point density and area density in the study area are 0.42 km -2 and 7.37% respectively. The maximum point density and maximum area density of landslide traces in the area are as high as 5.69 km -2 and 98.0% respectively. The landslides are primarily distributed in the relatively low-elevation northeastern part of Jianzha County, characterized mainly by large-scale loess landslides, with 14 landslides exceeding 1×10 6 m 2 . This inventory not only supplements the landslide trace data in the transition zone between the Qinghai-Tibet Plateau and the Loess Plateau, but also provides an important basis for subsequent landslide risk zoning, response to climate change, and landscape evolution. Additionally, it holds significant reference value for compiling landslide inventories in similar geological environments.

1 Introduction

Worldwide, mass movements such as landslides are prevalent geological hazards, causing heavy casualties ( Petley, 2012 ; Froude and Petley, 2018 ). As far as landslide hazards are concerned, China ranks among the regions with the very frequency of landslide hazards globally ( Kirschbaum et al., 2015 ; Xu and Xu, 2021 ). According to statistics from 2004 to 2016, China experienced 463 fatal landslides not induced by earthquakes, resulting in 4,718 deaths and causing economic losses exceeding 900 million dollars ( Zhang and Huang, 2018 ). Therefore, the prevention and control of landslide hazards is crucial for people’s lives. As a key step in hazard prevention and mitigation, including the analysis of regional landslide distribution patterns, hazard assessments, and risk assessment, the construction of a regional landslide inventory is fundamental and essential. A complete and accurate inventory ensures the objectivity and precision of subsequent work ( Xu, 2015 ; Piacentini et al., 2018 ).

In the construction of China’s landslide traces inventory, many scholars have carried out a lot of work and made certain progress ( Chen et al., 2016 ; Qiu et al., 2019 ; Zhao et al., 2019 ; Zhang et al., 2020 ). In northwestern China, Huang et al. (2022) compiled a landslide traces inventory for Hualong County, Qinghai Province, consisting of 3,517 landslides through visual interpretation of high-resolution optical images. Furthermore, an in-depth study on the spatial distribution patterns of landslides was conducted based on this inventory. In central China, Li et al. (2022a) primarily utilized visual interpretation, supplemented by existing literature and hazard records, to improve and supplement the landslide traces inventory for Baoji City, Shaanxi Province. The inventory contains a total of 3,422 landslides, providing foundational data for subsequent exploration of the distribution characteristics of large-scale landslides in the region. In the western part of the Qinghai-Tibet Plateau, Cui et al. (2023) employed the Google Earth platform and visual interpretation method to identify landslide traces in the Western Himalayan Syntaxis. They established a landslide traces inventory containing 7,947 landslides. This inventory serves as a support for subsequent landslide hazard assessments. Wu et al. (2016) collected landslide data based on aerial photographs at a scale of 1:50,000 under the conditions of existing data and field survey. They mapped 328 landslides in Gangu County, Gansu Province, providing a crucial foundation for subsequent research. Lan et al. (2004) combined aerial photographs, previous landslide investigation data, and on-site verification to compile a landslide inventory for the Xiaojiang River Basin, including 574 landslide records. They conducted spatial analysis and prediction of landslide based on this inventory. The landslide data sets constructed by these studies, supported by various methods, demonstrate the ability to facilitate subsequent study on landslide in terms of accuracy and completeness. Nevertheless, accurate and complete landslide trace data are still lacking for the entire region of China.

In studies covering Jianzha County, many scholars have carried out identification work on regional landslides, or conducted research on landslide failure patterns, InSAR deformation analysis, geomorphic effects, and other aspects based on landslide data ( Ma et al., 2008 ; Guo et al., 2020a ; Wang et al., 2022 ; Tu et al., 2023 ). Yin et al. (2014) primarily utilized visual interpretation to identify 508 landslides from Sigou Gorge to Lagan Gorge in the upper reaches of the Yellow River, with many landslides distributed in Jianzha County. Tu et al. (2023) conducted landslide detection in the upper reaches of the Yellow River based on InSAR technology, and carried out detailed deformation analysis of the Lijia Gorge landslides group in Jianzha County. Du et al. (2023) combined InSAR deformation monitoring and optical images to identify 597 landslides in the upper reaches of the Yellow River. Landslides are mainly distributed in Jianzha County and its surrounding areas. Wang et al. (2022) conducted deformation analysis on the Simencun landslide in Jianzha County to explore the relationship between the failure patterns before and after the landslide occurrence. Currently, although many studies have been carried out in Jianzha County based on landslide data. However, the landslide inventory maps produced do not cover the entire Jianzha County, or the landslide data are not complete and detailed enough. Therefore, by combining the visual interpretation of high-resolution optical images with the comparison of existing literature, this study compiled a landslide traces inventory for Jianzha County, Qinghai Province. Additionally, a spatial analysis was performed on the inventory. Finally, the completeness and importance of the landslide inventory are discussed.

2 Study area

Jianzha County has a total area of approximately 1714 km 2 and is located in the transitional zone from the upper reaches of the Yellow River on the northeastern edge of the Qinghai-Tibet Plateau to the Loess Plateau ( Figure 1 ) ( Ma et al., 2008 ). For a long time, the landscape evolution of this region has been influenced by the northeastward compression of the Qinghai-Tibet Plateau, resulting in the formation of basin and mountainous topography ( Guo et al., 2020a ; Peng et al., 2020 ). The overall terrain in the region is high in the southwest and low in the northeast. The northeastern part is the Qunke-Jianzha Basin, characterized by relatively low elevations and crossed by the main trunk of the Yellow River. On either side, there are two basins, namely, the Guide Basin and the Xunhua Basin. The Yellow River and its tributaries exert strong erosion and incision along the edges of the basins, with cutting depths exceeding 500 m. This has resulted in the formation of numerous erosion and accumulation terraces, as well as steep and rugged slopes, providing favorable conditions for landslide occurrence ( Craddock et al., 2010 ; Guo et al., 2020a ; Du et al., 2023 ).

Figure 1 . Location of the study area. Surface wave magnitude (Ms) is a measure of the strength of an earthquake, calculated from surface wave. The larger the value, the stronger the earthquake. Active fault data from Deng (2007) .

The study area exhibits undulating and rugged topography with well-developed valleys and gullies. The surrounding active tectonics are developed, with the north part of the area having the Lajishanbeiyuan Fault (LJSBYF) and the Lijishannanyuan Fault (LJSNYF). The NWW-SEE trending Daotanghe-Linxia Fault (DTH-LXF) and NNW-SSE trending Riyueshan Fault (RYSF) pass through the study area. Tectonic activity and climate change contribute to the frequent geological hazards ( Yin et al., 2014 ). The large, extra-large, and giant landslides in the region are typical and representative in China ( Guo et al., 2020b ; Yin et al., 2021 ). Some studies suggest that the tectonic uplift of the Qinghai-Tibet Plateau, as an internal dynamic factor, has led to the episodic incision of the Yellow River main and tributary channels, serving as the underlying cause for the formation of giant landslides ( Li et al., 2011 ). As shown in Figure 1 , there are several historical earthquakes with Ms greater than 5.0 around Jianzha County. The occurrence of landslides may be related to seismic activity or may be the result of landscape evolution, such as river erosion and high groundwater levels ( Guo et al., 2016 ; Guo et al., 2018 ).

With the advancement of remote sensing technology and improved transportation accessibility, the main methods for compiling regional landslide inventories currently include field investigation, visual interpretation of satellite images combined with computer, and automatic identification technology. Table 1 summarizes the advantages and disadvantages of the three landslide identification methods. Detailed field investigation can ensure high accuracy for landslide investigations in small-scale areas ( Huangfu et al., 2021 ). However, for large-scale regional landslide investigations, the feasibility of extensive field investigation decreases. This is primarily due to the substantial cost and time required ( Peng et al., 2016 ), as well as the difficulty in accessing rugged landslide sites. With the development of automatic identification technology, it has a significant advantage in quickly obtaining regional landslide data. However, its accuracy may be not very good ( Fayne et al., 2019 ; Zhang et al., 2020 ; Piralilou et al., 2021 ; Vecchiotti et al., 2021 ; Milledge et al., 2022 ). Combining the strengths of both approaches, the human-computer interaction visual interpretation of satellite images has gradually become an important method for constructing landslide inventory ( Xu et al., 2015 ; Shao et al., 2020 ; Li et al., 2021 ; Cui et al., 2022a ). This approach requires interpreters to have certain professional background knowledge. Compared to detailed field survey, it sacrifices a small portion of accuracy but significantly improves the efficiency of constructing landslide inventory ( Xu et al., 2014b ; Cui et al., 2022b ; Cui et al., 2022c ).

Table 1 . Advantages and disadvantages of three landslide identification methods.

This article primarily employed high-resolution optical images overlayed on terrain data for human-computer interactive visual interpretation, and combined existing landslide records in literature for validation and supplementation to construct a landslide traces inventory for Jianzha County. Google Earth Pro platform integrates a vast amount of high-resolution optical satellite image data and allows for the three-dimensional, multi-angle display of landscape by overlaying terrain data ( Crosby et al., 2012 ; Rabby and Li, 2019 ). This provides extremely convenient conditions for landslide identification. Focusing on the Jianzha County, the image quality is exceptionally high, with 100% satellite image coverage and 0% cloud coverage. Therefore, we performed repetitive basic work on landslide interpretation based on the Google Earth Pro for inventory construction. First of all, the shape and boundary of the landslide can be easily determined based on the differences between the texture, tone, shadow and vegetation development on the satellite images and the surrounding environment, combined with terrain differences and multi-angle observation. Secondly, many existing literature findings on landslides in the region will be conducted to check and supplement the inventory for ensuring the completeness and objectivity. Because different landslides have different topographic and geomorphic characteristics, there is no uniform standard applicable to the interpretation of all landslide traces. Here, some common landslide features used in landslide interpretation are listed: 1) Having an obvious armchair-shaped back wall and the phenomenon of double grooves homologous; 2) Depression in the source area, prominent topography in the accumulation area, accompanied by a distinct landslide boundary; 3) Obvious displacement between the landslide body and the surrounding environment, accompanied by cracks or differences in elevation; 4) The source area shows a brighter color, and the accumulation has transverse fissures and appears tongue-shaped; 5) Irregular stepped appearance in the accumulation body, with the terraces possibly transformed into residential areas or farmland.

4 Results and analysis

4.1 landslide traces inventory.

The landslide inventory serves as a crucial foundation for regional landslide risk assessment and prevention. Many scholars have conducted regional or individual landslide studies in Jianzha County ( Yin et al., 2014 ; Guo et al., 2020b ; Du et al., 2023 ; Tu et al., 2023 ). Although the study areas of these studies cover or partially cover Jianzha County, most have not established a complete landslide traces inventory that fully encompasses Jianzha County. Table 2 presents selected existing landslide records in Jianzha County. After objectively supplemented and validated by these records, the landslide inventory constructed in this study contains a total of 713 landslide traces ( Figure 2 ). The total area of these landslides is 134.46 km 2 . The total landslide area excluding the overlap area is 126.30 km 2 , accounting for 7.37% of the study area. The average landslide area is approximately 0.19 km 2 , with a minimum of 3,556 m 2 and a maximum of 11.13 km 2 . It can be found that landslides mainly occur on the slopes of the relatively low-elevation ridges in the northeastern part of Jianzha County. These landslides are widely distributed in towns such as Kanbula, Jiajia, Cuozhou, Maketang, and Angla, with a predominance of large-scale landslides. In the southwest, where the altitude is relatively high, landslides are sparsely distributed.

Table 2 . Selected recorded landslides in Jianzha County.

Figure 2 . Spatial distribution map of landslide traces.

4.2 Typical landslide display

In order to more intuitively display the landslides, several typical landslides were selected within the study area for display ( Figure 3 ). It can be found that the predominant landslide type is loess landslide. The landslide boundary is easily identified based on the discontinuity in texture and shape between the deposits and the surrounding environment. The material movement along the slope is evident. The displacement between the landslide deposits and the boundary visually demonstrates the movement direction and shape of the landslide. Over time, traces of human activity become visible on the deposits. After reconstruction, roads and buildings of various sizes are distributed across the deposits. These typical landslide examples can clearly capture the landslide morphology and material movement traces, which is of great value for the study of regional landslide failure mechanisms.

Figure 3 . Display of seven typical landslide traces in the study area.

4.3 Landslide density statistics

In order to quantitatively analyze the spatial distribution of landslides, landslide point density and area density are used to characterize the distribution and aggregation of landslides. After kernel density calculation with the search radius set to 2 km, the results are shown in Figure 4 . High point density areas are primarily concentrated in the northeastern part of the study area ( Figure 4A ), with the maximum density reaching 5.69 km -2 . This indicates that landslides in these areas are numerically dominant. The maximum landslide area density is 98.0% ( Figure 4B ). High area density areas are different from high point density areas in distribution. For instance, landslide area density is more significant relative to point density in areas close to the Kanbula Town. This indicates that the landslides in this area tend to be larger in scale.

Figure 4 . Landslide density map. (A) point density map; (B) area density map.

5 Discussion

5.1 landslide scale and the completeness analysis.

To explore the scale of landslides in Jianzha County, the cumulative landslide number was plotted against the landslide area in a double logarithmic coordinate system to show the relationship between them ( Figure 5 ). Where N represents the number of landslides exceeding a given area, A. It can be observed that the majority of landslides have a scale smaller than 1×10 6 m 2 , with only 14 landslides exceeding 1×10 6 m 2 in scale. The fitting formula for all landslides is l g N A = − 0.728   l g A + 5.957 , with R 2 =0.882. For landslides with an area larger than 1×10 5 m 2 , the fitting formula is l g N A = − 1.210   l g A + 8.552 , with R 2 =0.99, indicating that these data are relatively complete. For landslides with an area smaller than 1×10 4 m 2 , the curve exhibits a smoother trend, possibly due to the less distinct image change characteristics in the exposed loess areas, making them difficult to identify.

Figure 5 . Curve depicting correlation between the cumulative landslide number and the landslide area.

As shown in Figure 5 , landslides with an area greater than 1×10 5 m 2 are fitted as l g N A = − 1.210   l g A + 8.552 . In previous studies, this formula was often used in the statistics of coseismic landslide inventories to evaluate the completeness. For example, in the nearly complete coseismic landslide inventory established by Xu et al. (2014a) after the Wenchuan earthquake, landslides within a certain area are defined by the equation l g N A = − 2.0745   l g A + 13 , and the landslides exhibit a rolling trend. Similarly, coseismic landslide inventories for the Minxian-Zhangxian earthquake ( Xu et al., 2014b ) and Maerkang earthquake ( Chen et al., 2023 ) also show a similar trend, with the slopes and intercepts of the corresponding fitting equations are −1.341 and 6.02 (Minxian-Zhangxian earthquake) and −1.1052 and 5.7839 (Maerkang earthquake), respectively. Although the scale of landslides may vary due to different environmental conditions. However, it can be observed that whether it is the landslide traces inventory of this article or the coseismic landslide inventory, the landslides show a similar trend of change. In particular, in the work of establishing a landslide inventory with similar landslide scales, Li et al. (2022b) constructed a landslide traces inventory containing 3,757 landslides around the Baihetan Hydropower Station reservoir in China. The relationship between the cumulative number and area of landslides with an area greater than 1×10 5 m 2 is l g N A = − 1.275   l g A + 6.26 . Upon comparison, this is very close to the results of this article, which also proves the completeness of our inventory to a certain extent. By comparing the completeness of landslide inventories in different categories, it is concluded that landslide inventories of the same category have more reference value than those of different categories.

5.2 Objective assessment of methods

A complete and detailed landslide inventory is of great significance for regional landslide research and risk management. The human-computer interaction visual interpretation method, as one of the primary approaches for establishing regional landslide inventory, possesses advantages that are irreplaceable by field investigation and automatic identification techniques ( Guzzetti et al., 2012 ; Tian et al., 2019 ; Xu et al., 2020 ). While this study primarily relied on such a method to construct a relatively objective landslide traces inventory for Jianzha County, there are still some limitations. For small-scale landslides, due to the resolution limitations of satellite images, the coverage of topographic and geomorphic features, and the subjective factors from interpreters, it is inevitable that landslides with unclear identification characteristics will be missed. Compared with detailed field investigation, the visual interpretation method consumes less cost and time. Compared with automatic identification method, it is superior in accuracy and is currently a widely used method for identifying regional landslides ( Cui et al., 2021 ; Li et al., 2022a ; Sun et al., 2024 ). This method sacrifices some accuracy compared to field investigation, but greatly improves efficiency. Balancing the efficiency of automatic identification and the accuracy of field investigation is an exploratory and challenging task.

5.3 The importance of the landslide inventory

Landslide susceptibility refers to the probability of slope failure in a specific geological environment without considering triggering factors ( Akgun, 2012 ; Nikoobakht et al., 2022 ). As a fundamental component, landslide inventory plays an indispensable role in landslide susceptibility assessment. It provides essential information about landslides, including the number, scale, location. Based on landslide inventory, one can select a single assessment method and specific influencing factors for landslide susceptibility assessment ( Huangfu et al., 2021 ; Nanehkaran et al., 2021 ; Cemiloglu et al., 2023 ). Alternatively, one can choose several different assessment methods for comparative analysis to find the optimal results ( Azarafza et al., 2021 ; Nanehkaran et al., 2022 ; Mao et al., 2024 ). With the development of landslide assessment, machine learning has demonstrated outstanding performance among many methods, gradually becoming the preferred approach for assessment ( Nanehkaran et al., 2023 ). Based on susceptibility assessment, triggering factors are added to evaluate landslide hazard, while carrier indicators are added for vulnerability assessment. Risk assessment is then performed by overlaying hazard and vulnerability. However, regardless of which assessment method is chosen and which landslide influencing factors are considered, the susceptibility assessment, hazard assessment, vulnerability assessment, and risk assessment all need to be based on landslide data. The landslide inventory can not only be used to validate the results obtained through predictive modeling, but also provide an important reference for exploring factors involved in the occurrence of new landslides. Many studies have been carried out on incomplete landslide inventories and updated the inventories, effectively enhancing the understanding of subsequent landslide development and assessment research. For example, the Hokkaido earthquake ( Kasai and Yamada, 2019 ; Cui et al., 2021 ), Wenchuan earthquake ( Dai et al., 2011 ; Xu et al., 2014a ), and Jiuzhaigou earthquake ( Tian et al., 2019 ; Sun et al., 2024 ). The work of compiling a complete and detailed landslide inventory is not only of great value and significance, but also has important supporting value for subsequent research on landslide failure mechanisms, landscape evolution, especially landslide susceptibility assessment.

6 Conclusion

This study established a landslide traces inventory in Jianzha County, Qinghai Province, China, and conducted a statistical analysis of their number, area, and density. A total of 713 landslides were identified, mainly loess landslides. The total area of landslides is 134.46 km 2 , ranging in scale from 3556 m 2 to 11.13 km 2 . The landslides are primarily concentrated in the low-elevation regions of the northeastern part of the study area. This inventory is more similar in scale and completeness to other loess landslide inventories. Furthermore, it is more complete and detailed than previous landslide traces records in Jianzha County. The study compiled the most complete and detailed landslide traces inventory in Jianzha County so far, which is of great significance to landslide scientific research. In future, relevant research on loess landslide development characteristics, failure mechanisms, susceptibility assessment and risk zoning can be conducted based on this landslide inventory.

Data availability statement

The raw data supporting the conclusion of this article will be made available by the authors, without undue reservation.

Author contributions

TL: Investigation, Visualization, Writing–original draft. CX: Conceptualization, Resources, Writing–review and editing. LL: Investigation, Writing–review and editing. JX: Investigation, Writing–review and editing.

The author(s) declare financial support was received for the research, authorship, and/or publication of this article. This work was supported by the National Institute of Natural Hazards, Ministry of Emergency Management of China (2023-JBKY-57) and the National Natural Science Foundation of China (42077259).

Acknowledgments

Deep thanks are extended to the reviewers for their beneficial review and valuable comments.

Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The author(s) declared that they were an editorial board member of Frontiers, at the time of submission. This had no impact on the peer review process and the final decision.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Akgun, A. (2012). A comparison of landslide susceptibility maps produced by logistic regression, multi-criteria decision, and likelihood ratio methods: a case study at İzmir, Turkey. Landslides 9 (1), 93–106. doi:10.1007/s10346-011-0283-7

CrossRef Full Text | Google Scholar

Azarafza, M., Azarafza, M., Akgün, H., Atkinson, P. M., and Derakhshani, R. (2021). Deep learning-based landslide susceptibility mapping. Sci. Rep. 11 (1), 24112. doi:10.1038/s41598-021-03585-1

PubMed Abstract | CrossRef Full Text | Google Scholar

Cemiloglu, A., Zhu, L., Mohammednour, A. B., Azarafza, M., and Nanehkaran, Y. A. (2023). Landslide susceptibility assessment for Maragheh County, Iran, using the logistic regression algorithm. Land 12 (7), 1397. doi:10.3390/land12071397

Chen, W., Chai, H., Zhao, Z., Wang, Q., and Hong, H. (2016). Landslide susceptibility mapping based on GIS and support vector machine models for the Qianyang county, China. Environ. Earth Sci. 75, 474. doi:10.1007/s12665-015-5093-0

Chen, Z., Huang, Y., He, X., Shao, X., Li, L., Xu, C., et al. (2023). Landslides triggered by the 10 June 2022 Maerkang earthquake swarm, Sichuan, China: spatial distribution and tectonic significance. Landslides 20, 2155–2169. doi:10.1007/s10346-023-02080-0

Craddock, W. H., Kirby, E., Harkins, N. W., Zhang, H., Shi, X., and Liu, J. (2010). Rapid fluvial incision along the Yellow River during headward basin integration. Nat. Geosci. 3 (3), 209–213. doi:10.1038/ngeo777

Crosby, C. J., Whitmeyer, S. J., De Paor, D. G., Bailey, J., and Ornduff, T. (2012). Lidar and Google Earth: simplifying access to high-resolution topography data. Spec. Pap. Geol. Soc. Am. 492, 37–47. doi:10.1130/2012.2492(03)

Cui, Y., Bao, P., Xu, C., Ma, S., Zheng, J., and Fu, G. (2021). Landslides triggered by the 6 September 2018 Mw6.6 Hokkaido, Japan: an updated inventory and retrospective hazard assessment. Earth Sci. Inf. 14, 247–258. doi:10.1007/s12145-020-00544-8

Cui, Y., Hu, J., Xu, C., Miao, H., and Zheng, J. (2022b). Landslides triggered by the 1970 Ms7.7 Tonghai earthquake in Yunnan, China: an inventory, distribution characteristics, and tectonic significance. J. Mt. Sci. 19 (6), 1633–1649. doi:10.1007/s11629-022-7321-x

Cui, Y., Hu, J., Zheng, J., Fu, G., and Xu, C. (2022a). Susceptibility assessment of landslides caused by snowmelt in a typical loess area in the Yining County, Xinjiang, China. Q. J. Eng. Geol. Hydrogeology 55 (1), qjegh2021–2024. doi:10.1144/qjegh2021-024

Cui, Y., Jin, J., Huang, Q., Yuan, K., and Xu, C. (2022c). A data-driven model for spatial shallow landslide probability of occurrence due to a typhoon in Ningguo City, Anhui Province, China. Forests 13 (5), 732. doi:10.3390/f13050732

Cui, Y., Yang, W., Xu, C., and Wu, S. (2023). Distribution of ancient landslides and landslide hazard assessment in the Western Himalayan Syntaxis area. Front. Earth Sci. 11, 1135018. doi:10.3389/feart.2023.1135018

Dai, F., Xu, C., Yao, X., Xu, L., Tu, X., and Gong, Q. (2011). Spatial distribution of landslides triggered by the 2008 Ms8.0 Wenchuan earthquake, China. J. Asian Earth Sci. 40 (4), 883–895. doi:10.1016/j.jseaes.2010.04.010

Deng, Q. (2007). Map of active tectonics in China . Beijing: Seismological Press . (In Chinese)

Du, J., Li, Z., Song, C., Zhu, W., Ji, Y., Zhang, C., et al. (2023). InSAR-based active landslide detection and characterization along the upper reaches of the Yellow River. IEEE J. Sel. Top. Appl. Earth Observations Remote Sens. 16, 3819–3830. doi:10.1109/jstars.2023.3263003

Fayne, J. V., Ahamed, A., Roberts-Pierel, J., Rumsey, A. C., and Kirschbaum, D. (2019). Automated satellite-based landslide identification product for Nepal. Earth Interact. 23 (3), 1–21. doi:10.1175/ei-d-17-0022.1

Froude, M. J., and Petley, D. N. (2018). Global fatal landslide occurrence from 2004 to 2016. Nat. Hazards Earth Syst. Sci. 18 (8), 2161–2181. doi:10.5194/nhess-18-2161-2018

Guo, X., Forman, S. L., Marin, L., and Li, X. (2018). Assessing tectonic and climatic controls for late quaternary fluvial terraces in Guide, Jianzha, and Xunhua basins along the Yellow River on the northeastern Tibetan plateau. Quat. Sci. Rev. 195, 109–121. doi:10.1016/j.quascirev.2018.07.005

Guo, X., Sun, Z., Lai, Z., Lu, Y., and Li, X. (2016). Optical dating of landslide-dammed lake deposits in the upper Yellow River, Qinghai-Tibetan Plateau, China. Quat. Int. 392, 233–238. doi:10.1016/j.quaint.2015.06.021

Guo, X., Wei, J., Lu, Y., Song, Z., and Liu, H. (2020a). Geomorphic effects of a dammed pleistocene lake formed by landslides along the upper Yellow River. Water 12 (5), 1350. doi:10.3390/w12051350

Guo, X., Wei, J., Song, Z., Lai, Z., and Yu, L. (2020b). Optically stimulated luminescence chronology and geomorphic imprint of Xiazangtan landslide upon the upper Yellow River valley on the northeastern Tibetan Plateau. Geol. J. 55 (7), 5498–5507. doi:10.1002/gj.3754

Guzzetti, F., Mondini, A. C., Cardinali, M., Fiorucci, F., Santangelo, M., and Chang, K.-T. (2012). Landslide inventory maps: new tools for an old problem. Earth-Science Rev. 112 (1-2), 42–66. doi:10.1016/j.earscirev.2012.02.001

Huang, Y., Xu, C., Li, L., He, X., Cheng, J., Xu, X., et al. (2022). Inventory and spatial distribution of ancient landslides in Hualong county, China. Land 12 (1), 136. doi:10.3390/land12010136

Huangfu, W., Wu, W., Zhou, X., Lin, Z., Zhang, G., Chen, R., et al. (2021). Landslide geo-hazard risk mapping using logistic regression modeling in Guixi, Jiangxi, China. Sustainability 13 (9), 4830. doi:10.3390/su13094830

Kasai, M., and Yamada, T. (2019). Topographic effects on frequency-size distribution of landslides triggered by the Hokkaido Eastern Iburi Earthquake in 2018. Earth, Planets Space 71 (1), 89–12. doi:10.1186/s40623-019-1069-8

Kirschbaum, D., Stanley, T., and Zhou, Y. (2015). Spatial and temporal analysis of a global landslide catalog. Geomorphology 249, 4–15. doi:10.1016/j.geomorph.2015.03.016

Lan, H. X., Zhou, C. H., Wang, L. J., Zhang, H. Y., and Li, R. H. (2004). Landslide hazard spatial analysis and prediction using GIS in the Xiaojiang watershed, Yunnan, China. Eng. Geol. 76 (1), 109–128. doi:10.1016/j.enggeo.2004.06.009

Li, L., Xu, C., Xu, X., Zhang, Z., and Cheng, J. (2021). Inventory and distribution characteristics of large-scale landslides in Baoji city, Shaanxi province, China. ISPRS Int. J. Geo-Information 11 (1), 10. doi:10.3390/ijgi11010010

Li, L., Xu, C., Yang, Z., Zhang, Z., and Lv, M. (2022a). An inventory of large-scale landslides in Baoji city, Shaanxi province, China. Data 7 (8), 114. doi:10.3390/data7080114

Li, L., Xu, C., Yao, X., Shao, B., Ouyang, J., Zhang, Z., et al. (2022b). Large-scale landslides around the reservoir area of Baihetan hydropower station in Southwest China: analysis of the spatial distribution. Nat. Hazards Res. 2 (3), 218–229. doi:10.1016/j.nhres.2022.07.002

Li, X., Guo, X., and Li, W. (2011). Mechanism of giant landslides from Longyangxia valley to Liujiaxia valley along upper Yellow River. J. Eng. Geol. 19 (4), 516–529 [in Chinese, with English summary].

Google Scholar

Ma, X., Wang, L., Lv, B., and Ju, S. (2008). An investigation of geological hazards based on IRS-P 6 remote sensing data, Jianzha county, Qinghai province. Northwest. Geol. 41 (2), 93–100 [in Chinese, with English summary].

Mao, Y., Li, Y., Teng, F., Sabonchi, A. K. S., Azarafza, M., and Zhang, M. (2024). Utilizing hybrid machine learning and soft computing techniques for landslide susceptibility mapping in a Drainage Basin. Water 16 (3), 380. doi:10.3390/w16030380

Milledge, D. G., Bellugi, D. G., Watt, J., and Densmore, A. L. (2022). Automated determination of landslide locations after large trigger events: advantages and disadvantages compared to manual mapping. Nat. Hazards Earth Syst. Sci. 22 (2), 481–508. doi:10.5194/nhess-22-481-2022

Nanehkaran, Y. A., Chen, B., Cemiloglu, A., Chen, J., Anwar, S., Azarafza, M., et al. (2023). Riverside landslide susceptibility overview: leveraging artificial neural networks and machine learning in accordance with the United Nations (UN) sustainable development goals. Water 15 (15), 2707. doi:10.3390/w15152707

Nanehkaran, Y. A., Licai, Z., Chen, J., Azarafza, M., and Yimin, M. (2022). Application of artificial neural networks and geographic information system to provide hazard susceptibility maps for rockfall failures. Environ. Earth Sci. 81 (19), 475. doi:10.1007/s12665-022-10603-6

Nanehkaran, Y. A., Mao, Y., Azarafza, M., Kockar, M. K., and Zhu, H.-H. (2021). Fuzzy-based multiple decision method for landslide susceptibility and hazard assessment: a case study of Tabriz, Iran. Geomechanics Eng. 24 (5), 407–418. doi:10.12989/gae.2021.24.5.407

Nikoobakht, S., Azarafza, M., Akgün, H., and Derakhshani, R. (2022). Landslide susceptibility assessment by using convolutional neural network. Appl. Sci. 12 (12), 5992. doi:10.3390/app12125992

Peng, D., Xu, Q., Qi, X., Fan, X., Dong, X., Li, S., et al. (2016). Study on early recognition of loess landslides based on field investigation. Int. J. Georesources Environment-IJGE Former. Int'l J Geohazards Environ. 2 (2), 35–52. doi:10.15273/ijge.2016.02.006

Peng, J., Lan, H., Qian, H., Wang, W., Li, R., Li, Z., et al. (2020). Scientific research framework of livable Yellow River. J. Eng. Geol. 28 (2), 189–201 [in Chinese, with English summary].

Petley, D. (2012). Global patterns of loss of life from landslides. Geology 40 (10), 927–930. doi:10.1130/g33217.1

Piacentini, D., Troiani, F., Daniele, G., and Pizziolo, M. (2018). Historical geospatial database for landslide analysis: the catalogue of landslide occurrences in the Emilia-Romagna region (CLOCkER). Landslides 15 (4), 811–822. doi:10.1007/s10346-018-0962-8

Piralilou, S. T., Shahabi, H., and Pazur, R. (2021). Automatic landslide detection using bi-temporal sentinel 2 imagery. GI_Forum 9, 39–45. doi:10.1553/giscience2021_01_s39

Qiu, H., Cui, Y., Yang, D., Pei, Y., Hu, S., Ma, S., et al. (2019). Spatiotemporal distribution of nonseismic landslides during the last 22 years in Shaanxi province, China. ISPRS Int. J. Geo-Information 8 (11), 505. doi:10.3390/ijgi8110505

Rabby, Y. W., and Li, Y. (2019). An integrated approach to map landslides in Chittagong Hilly Areas, Bangladesh, using Google Earth and field mapping. Landslides 16 (3), 633–645. doi:10.1007/s10346-018-1107-9

Shao, X., Ma, S., Xu, C., Shen, L., and Lu, Y. (2020). Inventory, distribution and geometric characteristics of landslides in Baoshan city, Yunnan province, China. Sustainability 12 (6), 2433. doi:10.3390/su12062433

Sun, J., Shao, X., Feng, L., Xu, C., Huang, Y., and Yang, W. (2024). An essential update on the inventory of landslides triggered by the Jiuzhaigou Mw6. 5 earthquake in China on 8 August 2017, with their spatial distribution analyses. Heliyon 10 (2), e24787. doi:10.1016/j.heliyon.2024.e24787

Tian, Y., Xu, C., Ma, S., Xu, X., Wang, S., and Zhang, H. (2019). Inventory and spatial distribution of landslides triggered by the 8th August 2017 Mw 6.5 Jiuzhaigou earthquake, China. J. Earth Sci. 30 (1), 206–217. doi:10.1007/s12583-018-0869-2

Tu, K., Ye, S., Zou, J., Hua, C., and Guo, J. (2023). InSAR displacement with high-resolution optical remote sensing for the early detection and deformation analysis of active landslides in the upper Yellow River. Water 15 (4), 769. doi:10.3390/w15040769

Vecchiotti, F., Tilch, N., and Kociu, A. (2021). The use of TERRA-ASTER satellite for landslide detection. Geosciences 11 (6), 258. doi:10.3390/geosciences11060258

Wang, L., Qiu, H., Zhou, W., Zhu, Y., Liu, Z., Ma, S., et al. (2022). The post-failure spatiotemporal deformation of certain translational landslides may follow the pre-failure pattern. Remote Sens. 14 (10), 2333. doi:10.3390/rs14102333

Wu, Y., Li, W., Liu, P., Bai, H., Wang, Q., He, J., et al. (2016). Application of analytic hierarchy process model for landslide susceptibility mapping in the Gangu County, Gansu Province, China. Environ. Earth Sci. 75, 422. doi:10.1007/s12665-015-5194-9

Xu, C. (2015). Preparation of earthquake-triggered landslide inventory maps using remote sensing and GIS technologies: principles and case studies. Geosci. Front. 6 (6), 825–836. doi:10.1016/j.gsf.2014.03.004

Xu, C., Xu, X., and Shyu, J. B. H. (2015). Database and spatial distribution of landslides triggered by the Lushan, China Mw6.6 earthquake of 20 April 2013. Geomorphology 248, 77–92. doi:10.1016/j.geomorph.2015.07.002

Xu, C., Xu, X., Shyu, J. B. H., Zheng, W., and Min, W. (2014b). Landslides triggered by the 22 July 2013 Minxian–Zhangxian, China, Mw5.9 earthquake: inventory compiling and spatial distribution analysis. J. Asian Earth Sci. 92, 125–142. doi:10.1016/j.jseaes.2014.06.014

Xu, C., Xu, X., Yao, X., and Dai, F. (2014a). Three (nearly) complete inventories of landslides triggered by the May 12, 2008 Wenchuan Mw7.9 earthquake of China and their spatial distribution statistical analysis. Landslides 11, 441–461. doi:10.1007/s10346-013-0404-6

Xu, X., and Xu, C. (2021). Natural Hazards Research: an eternal subject of human survival and development. Nat. Hazards Res. 1 (1), 1–3. doi:10.1016/j.nhres.2020.12.003

Xu, Y., Allen, M. B., Zhang, W., Li, W., and He, H. (2020). Landslide characteristics in the Loess Plateau, northern China. Geomorphology 359, 107150. doi:10.1016/j.geomorph.2020.107150

Yin, Z., Qin, X., and Zhao, W. (2014). “Characteristics of landslides from sigou gorge to lagan gorge in the upper reaches of Yellow River. Landslide Science for a Safer Geoenvironment: vol. 1,” in The international programme on landslides (IPL) ( Springer ).

Yin, Z., Qin, X., Zhao, W., and Wei, G. (2013a). Characteristics of landslides in upper reaches of Yellow River with multiple data of remote sensing. J. Eng. Geol. 21 (5), 779–787 [in Chinese, with English summary].

Yin, Z., Wei, G., Qi, X., and Zhou, C. (2013b). Spatial and temporal characteristics of landslides and there response to climatic change from Sigou to Lagan gorges in upper reaches of Yellow River. J. Eng. Geol. 21 (1), 129–137 [in Chinese, with English summary].

Yin, Z., Wei, G., Qin, X., Li, W., and Zhao, W. (2021). Research progress on landslides and dammed lakes in the upper reaches of the Yellow River, northeastern Tibetan Plateau. Earth Sci. Front. 28 (2), 46–57 [in Chinese, with English summary].

Zhang, F., and Huang, X. (2018). Trend and spatiotemporal distribution of fatal landslides triggered by non-seismic effects in China. Landslides 15 (8), 1663–1674. doi:10.1007/s10346-018-1007-z

Zhang, P., Xu, C., Ma, S., Shao, X., Tian, Y., and Wen, B. (2020a). Automatic extraction of seismic landslides in large areas with complex environments based on deep learning: an example of the 2018 Iburi earthquake, Japan. Remote Sens. 12 (23), 3992. doi:10.3390/rs12233992

Zhang, Z., Wang, T., and Wu, S. (2020b). Distribution and features of landslides in the tianshui basin, northwest China. J. Mt. Sci. 17 (3), 686–708. doi:10.1007/s11629-019-5595-4

Zhao, B., Wang, Y., Chen, M., Luo, Y., Liang, R., and Li, J. (2019). Typical characteristics of large-scale landslides in the transition belt between the Qinghai-Tibet Plateau and the Loess Plateau. Arabian J. Geosciences 12, 470. doi:10.1007/s12517-019-4612-9

Keywords: landslide traces inventory, upper reaches of the yellow river, loess landslides, Jianzha County, visual interpretation

Citation: Li T, Xu C, Li L and Xu J (2024) The landslide traces inventory in the transition zone between the Qinghai-Tibet Plateau and the Loess Plateau: a case study of Jianzha County, China. Front. Earth Sci. 12:1370992. doi: 10.3389/feart.2024.1370992

Received: 15 January 2024; Accepted: 19 March 2024; Published: 05 April 2024.

Reviewed by:

Copyright © 2024 Li, Xu, Li and Xu. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

*Correspondence: Chong Xu, [email protected]

This article is part of the Research Topic

Prevention, Mitigation, and Relief of Compound and Chained Natural Hazards

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

• View all journals
• My Account Login
• Explore content
• About the journal
• Publish with us
• Sign up for alerts
• Open access
• Published: 10 February 2023

Case study of a landslide continuous probability rainfall threshold analysis based on the prediction interval principle

• Yu Huang 1 , 2 ,
• Cuizhu Zhao 1 ,
• Xiaoyan Jin 1 ,
• Yan Zhu 3 ,
• Ming Peng 1 &
• Zhiyi Chen 1  

Scientific Reports volume  13 , Article number:  2434 ( 2023 ) Cite this article

2241 Accesses

1 Altmetric

Metrics details

• Civil engineering
• Natural hazards

Bazhong City is located on stratum dominated by red-bed conditions. This type of weak geological condition with sand and mudstone interbedding is very easily affected by climatic conditions to produce rainfall-type landslides. On the basis of landslide data statistics collected in Bazhong City from 2011 to 2019, this paper uses ERA5-Land rainfall data to statistically analyze the correlation between rainfall and landslide events in Bazhong City. The landslide events in Bazhong City are greatly affected by rainfall events lasting for 10 days. Considering the influence of rainfall seepage and other processes, an effective cumulative rainfall-duration threshold curve for Bazhong City is obtained via median nonlinear fitting. Then, on the basis of the prediction interval, the rainfall threshold formula under different landslide occurrence probabilities is obtained and the critical threshold curve with a non-exceeding probability of 1% in Bazhong City is calculated and verified. Subsequently, a continuous probability distribution fitting function of landslide occurrence is established and a continuous probability distribution surface with a good fitting effect in Bazhong City is obtained. This allows a definite probability of whether future rainfall events will induce landslides to be obtained, providing an important basis for engineering disaster prevention and mitigation.

Similar content being viewed by others

Regional Rainfall Warning System for Landslides with Creep Deformation in Three Gorges using a Statistical Black Box Model

Zizheng Guo, Kunlong Yin, … Tengfei Wang

Dynamic landslide susceptibility analysis that combines rainfall period, accumulated rainfall, and geospatial information

Jae-Joon Lee, Moon-Soo Song, … Sang-Guk Yum

Spatial–temporal characterization of rainfall in Pakistan during the past half-century (1961–2020)

Ghaffar Ali, Muhammad Sajjad, … Hafiza Nayab Gul

Introduction

Landslides are among the most widely distributed, most influential, and most damaging geological disasters worldwide, and rainfall is one of the most important factors inducing landslides 1 . Rainfall not only increases the weight of the landslide mass but also reduces the shear strength of the slope, reducing the slope stability 2 . Currently, studies concerning the impact of rainfall on the landslide stability primarily focus on two aspects: conducting stability calculations or model tests on a certain slope 3 , 4 , 5 and establishing landslide rainfall thresholds based on regional historical landslide rainfall data statistics 6 , 7 , 8 .

Especially in mountainous areas, the number of slopes is large, the range is large, and data are lacking. Therefore, whether a rainfall event will induce a landslide is difficult to judge. Because of the large number of potentially unstable slopes, performing data measurements and monitoring for every slope are extremely difficult, resulting in a lack of analysis data. Consequently, an appropriate assessment method needs to be selected. Rainfall threshold research based on regional historical occurrences can provide a basis for landslide warnings. The most basic requirement is to have many years of landslide and rainfall data and then to analyze the statistical relationship between the occurrence of landslide events and the rainfall conditions to obtain the rainfall threshold that induces a landslide event 9 . The rainfall threshold is a critical rainfall parameter value that characterizes an induced landslide. This concept was first proposed by Caine 10 , who compiled 73 cases of rainfall-induced landslides and debris flows, characterized by the rainfall duration and the average rainfall intensity, and used the envelope under a scatterplot of the rainfall-induced landslides as the rainfall threshold. This is the now widely used rainfall intensity–duration threshold concept 10 , 11 . Of course, the current concept of rainfall thresholds also includes the cumulative rainfall–duration threshold 12 , 13 , 14 , cumulative rainfall–rainfall intensity threshold 15 , and total rainfall threshold based on rainfall-induced landslides.

A rainfall-induced landslide is a dynamic process that includes the gradual infiltration of rainfall and the gradual saturation of the rock and soil. In the case of a low rainfall amount with a long duration, the process of inducing landslides has a strong time series and a certain hysteresis. The impact of rainfall on landslide hazards includes two factors: rainfall duration and previous rainfall amount. Therefore, in a rainfall threshold analysis, the previous rainfall duration should first be determined. Studies have been conducted concerning the selection of the parameters for the previous rainfall duration, but there is no consensus at present. The length of the previous effective rainfall duration, as determined by different authors, varies greatly. For example, Wieczorek et al. determined a time of rainfall-induced slope instability of 5 days 16 , while Kim et al. established the impact of early rainfall as being up to 20 days 17 . The early rainfall is greatly affected by the selected range of the rainfall duration study; therefore, the length of the rainfall duration should be defined first in a threshold analysis.

The occurrence of rainfall-induced landslides has great uncertainty. A traditional threshold analysis can only roughly estimate whether a landslide can be induced in a certain region under certain rainfall conditions based on experience. Therefore, in the process of threshold calculation and analysis, the 5% non-exceeding probability rainfall threshold is usually selected 7 , 18 , 19 , 20 . However, miscalculations may be associated with this threshold. In current research, some studies have given the upper and lower limits of the rainfall threshold 21 , while others have calculated the rainfall thresholds of different probability levels 22 , 23 , 24 . The lower threshold refers to the record lower than that without a rainfall-induced landslide, which means that the slope will not fail until a certain rainfall intensity or rainfall amount is reached. The upper threshold refers to the limit of rainfall-induced landslide events, which both depends on extreme meteorological disasters and includes the failure of strongly stable slopes. Between these thresholds, there are different landslide occurrence probabilities that need to be quantified.

Bazhong City is in the northeastern part of the Sichuan Basin, where mountains and hills are widely distributed, accounting for 98.2% of the total area. The terrain in the region varies greatly, the stratum and lithology are complex, and there are developed fold structures. The climate in this region is complex and changeable, and the geological disasters are multi-faceted, with strong concealment, significant danger, and considerable potential damage. According to statistics from the Bazhong Municipal Bureau of Land and Resources, by the end of 2014, there were 4432 hidden hazard points that could threaten the lives and property of residents in Bazhong City. These identified geological hazards primarily include landslides, debris flows, collapses, and potentially unstable slopes. Landslide geological disasters constituted the majority. According to statistics, there are 4014 landslide hazards, accounting for 90.6% of the total 25 . The density of hidden hazards in Bazhong City ranks second in Sichuan Province, and the number of people threatened per unit area ranks first in the province, with a maximum of 16.5 people threatened per square kilometer, which is approximately four times the average in Sichuan Province. This means that the lives of more than 240,000 people and property valued at nearly 5.4 billion yuan are threatened to varying degrees. One of the reasons for these hidden hazards is that the lithology of Bazhong City consists of primarily Triassic–Jurassic sand–mudstone interbeds 26 . This is a typical soft and hard red-bed rock mass, having characteristics of low strength, easy weathering, poor hydraulic properties, and large deformation. At the same time, the water resistance of mudstone and shale causes small particles carried by infiltrating rainwater to slowly discharge along the interface with the sand layer. Under the action of dry and wet cycles, weak siltized intercalation is formed, which can easily develop into an interlayer shear zone, causing slope instability. As a result of this special geological condition, large numbers of landslides occur in Bazhong City under rainfall conditions. This has a significant impact on the lives of residents. According to existing research 26 , landslide hazards in Bazhong City are primarily small and shallow soil landslides, which are highly sensitive to rainfall. Landslide areas are generally in a basically stable state under natural conditions and are prone to slip deformation under heavy rainfall conditions. Landslides in this region have small sizes, occur in large quantities, have high density, are prone to failure, and have an uneven spatial and temporal distribution.

The “September 16 torrential rain event” in 2011 triggered a large number landslides in Bazhong City, Nanjiang County, reaching a total of 1162 slides; this event inspired multiple studies concerning the rainfall threshold in red-bed areas. Taking the Erhuangping landslide in Nanjiang County as an example, Lu et al. analyzed the impact of rainfall on soil landslides in a red-bed area and obtained a fitting formula for the relationship between the rainfall intensity and the duration 27 . At the same time, in the process of establishing a prediction model for large-displacement landslides induced by heavy rainfall, Yu et al. produced two rainfall threshold curves for Nanjiang County 28 . In addition, on the basis of the rainfall duration and rainfall process, Zhang et al. discussed the rainfall types that trigger landslides in Bazhong City and found that such rainfall types were primarily incremental and unimodal 29 . However, there is no specific research concerning the rainfall threshold conditions for landslide occurrences in Bazhong City.

From June 19 to 20, 2019, Bazhong City was again hit by a heavy rainfall event. The maximum rainfall was 196.1 mm (from 20:00 on June 19 to 7:00 on June 20) in Guangong Town, Enyang District. The heavy rainfall caused floods in Bazhong City and greatly elevated the risk of landslides on the local mountains and slopes. Following the heavy rainfall, a large number of statistical investigations concerning landslide incidents in Bazhong City were performed. A total of 165 slope failures were identified from June to July. This event was sufficient to draw attention to rainfall-induced landslide events in Bazhong City. Because of differences in the geotechnical properties, slope gradients, and deterioration degrees of slopes, there will always be differences in whether and when landslides are induced on a slope given the same type of rainfall event. Therefore, it is necessary to perform threshold analyses using probability levels and statistics. In particular, obtaining the threshold curve under a certain non-exceeding probability and the landslide probability of different rainfall events is of great significance to provide early warnings of slope engineering disasters in this region.

Here, we study the rainfall threshold that triggers landslides in the area of Bazhong City, Sichuan Province, China, which is prone to rainfall-type landslides. By statistically analyzing the collected rainfall and landslide data from 2011 to 2019, a calculation formula of the threshold curve under different probabilities can be obtained. We then calculate and verify the threshold under a non-exceeding probability of 1%. Accordingly, a fitting formula for the continuous probability distribution of the landslide events is established. Using the effective cumulative rainfall and the previous rainfall duration, the probability of landslide events can be evaluated over time, providing a basis for forecasts and early warnings of slope geological disasters in Bazhong City.

Study area and materials

Bazhong City is located in the northeastern part of Sichuan Province, south of the boundary between the Qinling Mountains and the Huaihe River in China, bordering Dazhou in the east, Nanchong in the south, Guangyuan in the west, and Hanzhong in Shaanxi in the north. Its specific geographical location is shown in Fig.  1 , between 106° 20′ E and 107° 49′ E longitude and between 31° 15′ N and 32° 45′ N latitude, with an area of 12,300 km 2 . Bazhong City has a total of five county-level administrative divisions, including two municipal districts and three counties.

Elevation Map of Bazhong City, Sichuan Province, China.

Bazhong City is located in a typical mountainous area, with high terrain in the north and low terrain in the south, as shown in Fig.  1 . The terrain has a minimum elevation of 268.3 m in Huangmeixi, Pingchang County, in the south. The highest altitude is 2507 m in Guangwu Mountain, Nanjiang County, in the northwest. There are many mountains and hills in the study area; therefore, the area includes flowing water erosion, sedimentation, fan-shaped landform and gravity accumulation, residual landforms, and karst landforms. Bazhong City belongs to a typical subtropical humid monsoon climate zone, with abundant rainfall and a mild climate. According to the Bazhong City Yearbook, the average precipitation from 2011 to 2019 was 1229 mm. The annual rainfall in Nanjiang County was the largest in 2011 at 1910.1 mm, and the annual rainfall in Enyang District in 2015 was the smallest at 838.5 mm. The average annual rainfall distribution calculated based on the rainfall data from 2011 to 2019 is shown in Fig.  2 , which shows a decrease from southeast to northwest resulting from the combined influence of the monsoon climate and the topography.

Average annual rainfall distribution from 2011 to 2019.

This study collected landslide hazard data in Bazhong City from 2011 to 2019. Part of the data comes from the Natural Resources and Planning Bureau of Bazhong City, Sichuan Province, and another part comes from the collection of thesis materials 30 . A total of 515 landslide hazard events were collected in five counties and districts over 9 years; the distribution of these events is shown in Fig.  3 . In this study, only the approximate point of landslide occurrence is obtained, regardless of its influence range. According to the landslide data statistics, we primarily considered several situations, such as slope instabilities caused by rainfall, induced landslides or debris flows, and further aggravated landslides. In addition, these landslide events pose potential harm to human engineering activities, or have caused harm, including catastrophic landslide events and slope instability events that have not yet developed into disasters. The point density in Fig.  3 is calculated using the kernel density estimation method 31 . The data in the legend represent the number of landslide points per square kilometer.

Distribution map of landslide hazard points in Bazhong City from 2011 to 2019.

The acquisition of rainfall data includes ground station observations, radar observations, and satellite remote-sensing observations. Ground-observed rainfall data are limited by the number of stations established. Radar-estimated precipitation is affected by radar wave attenuation and complex terrain occlusion, and the obtained temporal and spatial coverage is limited. The ERA5 land precipitation data used in this paper have a resolution of 0.1° × 0.1°. A physical model based on global observations was used for the reanalysis. These climate data have high accuracy. The ability of the dataset to capture precipitation events is good; this has certain advantages for the daily-scale analysis required in this study and, most importantly, makes it easy to obtain data 32 , 33 . However, no matter how high the precision of the raster data, values can only be provided on a certain latitude and longitude grid 34 , 35 . In the process of extracting rainfall data corresponding to landslide events at specific spatial locations and time points, spatial linear interpolation was further adopted, ignoring the non-stationary variability of the rainfall in the spatial and temporal distribution 36 .

Statistical analysis

According to the Bazhong City Yearbook, statistics were compiled for the average monthly rainfall and the occurrence time of landslide hazards over many years, as shown in Fig.  4 . Landslide hazards primarily occur from May to October, with the number of disasters generally increasing with increasing rainfall and decreasing with decreasing rainfall, indicating that there is a close relationship between the two.

Monthly rainfall and landslide hazard frequency from 2011 to 2019.

To illustrate the relationship between landslide hazards and daily rainfall, this study performed frequency statistics of landslide hazards according to the rainfall grades shown in Table 1 37 . In addition, even though a total of 515 landslide hazards were identified in Bazhong City from 2011 to 2019, many landslide events occurred on the same day. To better explain the relationship between landslide hazards and daily rainfall, this study de-duplicated the number of landslides that occurred on the same day; that is, all landslide events that occurred during one day were recorded as a single disaster day.

Judging from the number of landslide days, it can be seen that, when the daily rainfall exceeds 50 mm, the number of disaster days increases. The data indicate that landslide hazards are more likely to occur with heavy rainfall weather. However, the collected landslide events indicate that the frequency of landslide hazards is approximately 50% when there is no rain or light rain on the disaster day. There are two reasons why the analysis produces this result. The first reason is that the collected landslide events are not induced by rainfall factors or that they are potential landslide events that have been in a state of creep. It is necessary to screen landslide events on the basis of a comprehensive analysis of rainfall events and landslides to obtain the rainfall-type landslide events. The second reason is that the accumulation of precipitation and seepage has a significant influence on the induced landslide events. This reason is very important; the concepts of rainfall intensity duration threshold, cumulative rainfall intensity threshold, and total rainfall threshold emphasize the impact of slope instabilities caused by early rainfall accumulation.

Statistics were acquired with respect to the grades of the average daily rainfall in the previous stage of the induced landslides, as shown in Fig.  5 . It can be seen that, when the previous rainfall lasted for 5 days, the daily average rainfall was primarily heavy rain. However, the rainfall events that triggered landslides in Bazhong City were primarily moderate rainfall events lasting 6–10 days. It can be seen from this that the landslides in Bazhong City have a strong relationship with the early accumulated rainfall. When calculating the rainfall threshold, at least 10 days of rainfall duration are taken for the analysis. Therefore, in terms of the subsequent selection of the rainfall threshold, we counted and analyzed the rainfall data 15 days prior to the landslide event. This includes most landslide events induced by continuous rainfall and reduces the statistical workload of the rainfall data to a certain extent.

Relationship between the average daily rainfall in the previous period and the frequency of landslides.

To obtain the probability of a slope instability landslide under a certain rainfall condition, we propose a further fitting method to obtain the continuous probability threshold surface based on existing rainfall threshold research. The main process is as follows.

Using the previous data statistics and analysis, select the appropriate rainfall threshold model. In this study, the importance of early rainfall is evident, so the cumulative rainfall duration threshold model is adopted.

Perform a power law function nonlinear fitting of the rainfall threshold. At the same time, calculate the standard deviation of the residual difference and statistically analyze its distribution.

On the basis of the prediction interval principle, the rainfall threshold under any non-exceedance probability is obtained through the inverse operation of the cumulative distribution function. For residuals that do not conform to normal distribution, the normal distribution should be converted.

By fitting the threshold surface, the calculation formula of the landslide occurrence probability under arbitrary rainfall can be obtained.

Rainfall threshold

The frequency of landslide hazards does not always increase with increasing rainfall duration. That is, in a long-term rainfall process, not all rainfall impacts the occurrence of landslides. In the process of rainfall, there are effects such as runoff, discharge, and evaporation, especially with increasing duration, and these effects cannot be ignored. To overcome this problem, Crozier proposed using the effective cumulative rainfall ( \({E}_{\mathrm{L}}\) ) to represent the impact of cumulative rainfall events; this method is widely used 38 , 39 :

In Eq. ( 1 ), \({E}_{\mathrm{L}}(0)\) is the rainfall on the day of the landslide; \({E}_{\mathrm{L}}(i)\) is the daily rainfall on the i th day prior to the landslide; \(N\) is the number of days of previous rainfall; and \(k\) is the effective rainfall coefficient, whose value ranges from 0.6 to 0.93 40 , 41 , depending primarily on the local rainfall permeability. In most studies, a value of k of 0.8 is used 17 , 42 , 43 , 44 . Because the occurrence of landslides in Bazhong City is primarily affected by rainfall events within 10 days and the influence of rainfall prior to 11–15 days is ignored, the k value in this study was taken to be 0.84 38 , 45 , 46 . At this value, 17% of the precipitation on the 10th day contributes to the occurrence of landslides.

On the basis of the formula of the effective cumulative rainfall, the relationship between the effective cumulative rainfall and the rainfall duration was derived 14 ; that is, the model shown in Eq. ( 2 ) is the rainfall threshold in Bazhong City. The rainfall duration was taken as the period from the first day of the previous rainfall duration to the day when the landslide occurred; the analysis was performed in units of days.

In Eq. ( 2 ), \(\mathrm{\alpha }\) is the scaling constant, which is the intercept of the threshold curve; \(\upgamma\) is the shape parameter, which defines the slope of the power function curve; and \(\Delta \mathrm{\alpha }\) and \(\Delta\upgamma\) represent the uncertainties in \(\mathrm{\alpha }\) and \(\upgamma\) , respectively 47 . This equation can be directly obtained via nonlinear fitting and can provide the fitting parameter range under a certain confidence level.

Prediction interval principle

In previous analyses, to obtain the landslide threshold under a certain non-exceeding probability, the frequentist method has usually been used 48 . Similarly, the principle of the prediction interval can provide predictions according to certain methods and rules based on the collected rainfall-induced landslide events, so as to determine the occurrence or probability of landslides in advance. Assuming that the prediction errors are normally distributed, the prediction interval at a confidence level of \(1-2\alpha\) is

In Eq. ( 3 ), \(\widehat{y}\) represents the fitted value for a given independent variable x in the regression analysis; \(\sigma\) is the standard deviation of the residual obtained in the process of fitting the power function; and \(Z\) represents the distance from the mean to zero with a probability of \(\alpha\) or 1 − \(\alpha\) in the standard normal distribution. According to the empirical value of a normal distribution, when \(\alpha =5\mathrm{\%}\) , \(\mathrm{Z}=-1.64\) and, when \(\alpha =95\mathrm{\%}\) , \(\mathrm{Z}=1.64\) .

To obtain the threshold curve under a certain non-exceeding probability in a landslide-induced rainfall threshold analysis, it is first necessary to obtain the \(Z\) value. This can be obtained via the inverse cumulative distribution function of a normal distribution, which is \({Z}_{\alpha }={\Phi }^{-1}(\alpha )\) . The cumulative distribution function of a normal distribution is

In Eq. ( 4 ), \(\mu\) is the mean value of the residuals in the regression fit. The inverse function of this equation is defined in many programs, for example, it can be obtained via the “norminv” function in the MATLAB software.

In addition, for a function whose fitting residuals do not conform to a normal distribution, function conversion is required. For the fitting of the rainfall threshold, the power law function is generally used and the residuals generally conform to a lognormal distribution. Accordingly, Eq. ( 3 ), based on the prediction interval principle, was modified to obtain a non-exceeding probability calculation formula that conforms to the rainfall threshold analysis:

In Eq. ( 5 ), \({\sigma }_{lgR}\) is the standard deviation of the logarithmic residual. The \({Z}_{\alpha }{\sigma }_{lgR}\) component can be directly expressed as the inverse function \({\Phi }^{-1}(\alpha )\) of a normal distribution, and the normal distribution \(\Phi \left(z\right)\) obeys the logarithmic residual mean and standard deviation distribution. In this context, in the rainfall threshold analysis, the mean and standard deviation obeyed by the logarithmic residual need to be obtained first via a nonlinear fitting of the median. Then, the threshold curve under any non-exceeding probability can be obtained using Eq. ( 5 ).

Continuous probability distribution model

To more accurately and conveniently prevent future rainfall-type landslide events in Bazhong City, it is necessary to analyze the continuous probability of landslide events in this region. Because the probability distribution function is a sigmoid function, this study establishes the following binary function to characterize the effect of the effective rainfall accumulation and duration on the probability of landslide occurrence:

In this formula, \(x\) and \(y\) are independent variables representing the effective accumulation of rainfall and the rainfall duration, respectively; and \(a\) , \(b\) , and \(c\) are fitting parameters.

Results and validation

According to the analysis results of the aforementioned landslide hazards and rainfall data, landslide hazards in Bazhong City are dominated by moderate rain lasting approximately 10 days. The landslide rainfall data were rectified considering the influence of non-rainfall-induced landslides. Data for landslides induced by 15 days of continuous rainfall in the early stage were not included because it is unclear whether the rainfall lasted for 15 days or longer; that is, this part of the data was not sufficiently complete, which has a certain impact on the analysis results. Considering the effective rainfall and the previous rainfall duration, a total of 391 landslide rainfall disaster records were screened. In this study, rainfall-induced landslide events were divided into two datasets. One portion of the data was used for threshold curve fitting, and the other portion was used to validate the threshold results. This study used landslide data from January 2011 to August 2019, with a total of 329 records, to calculate the rainfall threshold. At the same time, to calibrate the threshold, we counted 775 non-landslide points during the period from January 2011 to June 2019 when no landslides were induced.

Landslide rainfall threshold

Median nonlinear fitting based on a power function was performed on the 329 screened records, and the results are shown in Fig.  6 . The distribution of rainfall-type landslide events in Bazhong City is relatively discrete, and the fitting effect is not very good. Therefore, considering the uncertainty of the fitting, the fitting confidence interval with a confidence level of 95% and the confidence ranges of the coefficients \(\mathrm{\alpha }\) and \(\upgamma\) were calculated. The confidence intervals of the fitted results are shown in the shaded region in Fig.  6 . Considering the uncertainty of the fitting coefficients \(\mathrm{\alpha }\) and \(\upgamma\) , the median nonlinear fitting threshold curve and the standard deviation of the fitting parameters are expressed as

Landslide rainfall threshold curve for Bazhong City.

Figure  6 shows that the use of a 1% non-exceeding probability threshold has a poor early warning effect, with 107 landslide rainfall events not triggered. However, compared with the rainfall threshold fitted by the median line (50% non-exceeding probability threshold), it has a good early warning effect and only 18 heavy rainfall events did not induce landslides. This comparison indicates that the rainfall threshold with nonlinear fitting has a better warning effect.

To illustrate the validity of the rainfall threshold curve obtained by the nonlinear regression analysis, a residual analysis was performed on the actual effective cumulative rainfall value and the value obtained via fitting. Because this study uses a power function relationship to characterize the rainfall duration and the effective cumulative rainfall that induces landslides, the frequency distribution of the logarithmic residuals was calculated, as shown in Fig.  7 . This distribution basically conforms to a normal distribution with a mean of − 0.0221 and a standard deviation of 0.13601. This indicates that the rainfall events that induce the Bazhong landslides are discretely distributed above and below the fitted curve in the form of a log-normal distribution.

Logarithmic residual distribution of the rainfall threshold curve fitting for Bazhong City.

Figure  7 clearly shows that the logarithmic residuals are normally distributed. This indicates that it is reasonable to calculate the threshold under any subsequent non-exceedance probability using Eq. ( 5 ). On this basis, we performed threshold calculations under different non-exceedance probabilities, providing a data basis for subsequent fitting of the threshold surface.

Continuous probability distribution function

Using the rainfall-induced landslide data from January 2011 to August 2019, the probability of a landslide occurrence caused by a rainfall event with continuous rainfall for more than 2 days can be determined using Eqs. ( 2 ) and ( 5 ). A three-dimensional surface fitting can then be performed using Eq. ( 6 ) to obtain the threshold surface of the rainfall-triggered landslide events under the continuous distribution probability for Bazhong City, as shown in Fig.  10 a. The fitting results and the standard deviation of the fitting parameters are given in Table 2 . The goodness of fit is 0.999, which is approximately equal to 1. This shows the superiority of using Eq. ( 6 ) for the probability distribution fitting and indicates that good fitting results are obtained.

Verification

In Bazhong City in 2019, following the peak landslide frequency caused by the flood in June and under the influence of the continuous rainfall conditions, landslide events occurred one after another at different times. In particular, during two rainfall events in September and October, the frequency of landslides again increased. Both rainfall events lasted for approximately 10 days. The two rainfall events together caused more than 60 landslides. See Fig.  8 for details.

Two persistent rainfall-induced landslide events in September and October 2019.

The threshold curve proposed in this study was verified using the rainfall-induced landslide data from September to October. As shown in Fig.  9 , the rainfall events that induced landslides during these 2 months were discretely distributed above and below the nonlinear median fitting curve and the average logarithmic residual error of this part of the data was calculated to be − 0.07. In addition, this part of the data is above the 1% non-exceeding probability threshold curve. This illustrates the validity of the threshold curve obtained in this study.

Rainfall threshold validation.

At the same time, according to the logarithmic residual value of this part of the data and the nonlinear fitting, the probability of each rainfall-induced landslide event in September and October 2019 was obtained, as shown in Fig.  10 a. Compared with the fitting result of Eq. ( 6 ), the root mean square error is − 0.029, which shows the validity of the fitting result. In addition, Fig.  10 b shows the landslide probability induced by this part of the rainfall event and the residual of the fitting surface. Interestingly, the residual change in the figure also shows an s-shape; however, there is fortunately no large residual value.

( a ) Continuous probability distribution surface of rainfall-type landslides in Bazhong City and ( b ) residuals.

This paper describes a comprehensive method for establishing a continuous probability rainfall threshold based on the prediction interval principle. Several key issues concerning the workflow are discussed here.

Impact of early rainfall

In this study, first, the relationships between the landslide events in Bazhong City and the rainfall intensity on the day of the event and the duration of the previous rainfall are analyzed. Table 1 and Fig.  5 clearly show that the rainfall inducing landslides in Bazhong City is mostly moderate rain lasting for 6–10 days, while light rain is very likely to occur on the day of a landslide. This emphasizes the important impact of early rainfall on landslide induction. Considering the discharge and evaporation processes of early rainfall, most studies focus on the effective accumulated rainfall. However, the selection of the effective rainfall coefficient in Eq. ( 1 ) is subjective. Additional scientific methods are needed to evaluate the rainfall infiltration process and obtain the appropriate effective rainfall coefficient. Especially in this study, when selecting the effective rainfall coefficient, considering that the constant value of relevant studies is 0.8 17 , 42 , 43 , 44 and that the rainfall 6–10 days before a landslide in Bazhong City has a great impact, a greater value of 0.84 is used 38 , 45 , 46 . At present, this can be considered using complex decision-making algorithms 49 . At the same time, many studies have focused on the increase in soil moisture caused by early rainfall 50 , 51 . However, the simplest and most direct method is to establish a three-dimensional rainfall threshold model 52 . Through a simple and intuitive method, such a model can show the impact of early rainfall on landslide occurrence and can be based on threshold division, which has good feasibility and early warning performance.

Uncertainty based on rainfall threshold warning

Figure  6 indicates that using a certain rainfall threshold curve with different non-exceeding probabilities has different prediction effects on whether rainfall events induce landslides. The adoption of a rainfall threshold with a low non-exceedance probability will lead to seriously inefficient identification of rainfall landslides. Moreover, most landslides cannot be effectively identified using a high non-exceedance probability rainfall threshold. However, in the verification of subsequent rainfall-triggered landslide events (Fig.  9 ), all landslide events occurred on the 1% non-exceeding probability threshold curve, while only one-third of landslides occurred on the 50% non-exceeding probability threshold. This makes it difficult to choose which of the two is more suitable as the rainfall threshold for landslides in Bazhong City. Because it considers the influences of many possible factors, such as vegetation conditions, slope, elevation, and weathering degree, the rainfall threshold has great uncertainty. This problem has also been noted in previous studies 21 , 24 , but further research has not yielded a threshold function of the continuous distribution probability. Given a certain rainfall condition, the probability of landslide occurrence can be directly obtained, which is more suitable for project decision-making.

Comparing Figs.  6 and 9 , the selection of a rainfall threshold for inducing landslides should clearly be made based on an extensive collection of local rainfall events; this should not only prevent high false alarm rates but also improve the identification of landslide events. Some relatively good related research work has been carried out; for example, Segoni et al. developed the MaCumBA software to obtain the rainfall threshold with the smallest false alarm rate through multiple operations and reverse analyses 53 . Such methods should be an indispensable component of future rainfall threshold research.

Further study

There are still some deficiencies in the process of carrying out the above research, and future engineering early warning research could be carried out based on the following.

The accuracy of analyzing the probability of rainfall-induced landslides in a certain area using historical empirical data depends largely on the collection of sample data. Compared with other studies with thousands of landslide records, this study warrants further data collection and records. Only by constantly expanding the sample data and updating the probability threshold of rainfall landslides can geological disaster early warning work be improved.

To obtain a more scientific rainfall threshold, we need to consider a large number of rainfall events, including those that triggered landslides and those that did not. On the basis of this, landslide events can be effectively identified to minimize the rate of misjudgment. Therefore, further study is necessary to obtain the lowest false alarm rate based on continuous probability threshold analyses.

In the same area, even though there are similar stratigraphic conditions, conditions such as the slope, elevation, and vegetation cover differ, which will lead to uncertainties in the rainfall threshold. Risk analyses of certain areas should be performed by obtaining continuous probability rainfall thresholds combined with landslide susceptibility analyses. At the same time, the landslide warning effect can gradually be improved by combining real-time or climate prediction technologies.

Conclusions

On the basis of the collection of landslide hazards and ERA5-Land high-resolution rainfall data in Bazhong City from 2011 to 2019, this study used the concept of effective cumulative rainfall to conduct a refined rainfall threshold analysis of landslide hazards in Bazhong City. The main conclusions are as follows.

Landslide hazards in Bazhong City primarily occur in the period of June–October when the monthly average rainfall is high; events are especially concentrated in the period of June–July. Moderate rain-induced landslide hazards lasting approximately 10 days are the main type of event, accounting for approximately 70% of the collected data.

A power function fitting of the rainfall and landslide hazard events in Bazhong City was performed via nonlinear regression, and a threshold curve of \({E}_{\mathrm{L}}=98.05{\mathrm{D}}^{0.1166}\) was obtained. The threshold curve under the 1% non-exceeding probability is \({E}_{L-0.01}=44.97{\mathrm{D}}^{0.1166}\) . The threshold results were verified against two sustained rainfall events in September and October 2019, and the rainfall events were all above the 1% threshold line.

This study proposes using the continuous probability rainfall threshold surface for quantitative analyses. On the basis of the Sigmoid function, a continuous probability distribution function of rainfall-induced landslides was established and good fitting results were obtained. Using this method, we can preliminarily solve the problem of selecting an appropriate non-exceeding probability rainfall threshold.

The above research results provide a powerful tool to aid in the early warning of rainfall-induced landslides in Bazhong City. As a typical regional case prone to rainfall-induced landslides, the study of Bazhong City can provide a reference for studies of rainfall-induced landslides in other regions.

Data availability

The data that support the findings of this study are available from the corresponding author upon request. We used publicly available terrain data ( https://earthexplorer.usgs.gov/ ) and rainfall data from the European Centre for Medium-Range Weather Forecasts ( https://www.ecmwf.int/en/era5-land ).

Pereira, S., Zezere, J. L., Quaresma, I. D. & Bateira, C. Landslide incidence in the North of Portugal: Analysis of a historical landslide database based on press releases and technical reports. Geomorphology 214 , 514–525 (2014).

Article   ADS   Google Scholar  

Lepore, C., Arnone, E., Noto, L. V., Sivandran, G. & Bras, R. L. Physically based modeling of rainfall-triggered landslides: A case study in the Luquillo forest, Puerto Rico. Hydrol. Earth Syst. Sci. 17 , 3371–3387 (2013).

Tang, G., Huang, J., Sheng, D. & Sloan, S. W. Stability analysis of unsaturated soil slopes under random rainfall patterns. Eng. Geol. 245 , 322–332 (2018).

Article   Google Scholar  

Gallage, C., Abeykoon, T. & Uchimura, T. Instrumented model slopes to investigate the effects of slope inclination on rainfall-induced landslides. Soils Found. 61 , 160–174 (2021).

Shi, Z.-N. et al. A study of water content distribution and shallow stability of earth slopes subject to rainfall infiltration. Rock Soil Mech. 41 , 980 (2020).

Google Scholar  

Harilal, G. T., Madhu, D., Ramesh, M. V. & Pullarkatt, D. Towards establishing rainfall thresholds for a real-time landslide early warning system in Sikkim, India. Landslides 16 , 2395–2408 (2019).

Peruccacci, S. et al. Rainfall thresholds for possible landslide occurrence in Italy. Geomorphology 290 , 39–57 (2017).

Nolasco-Javier, D., Kumar, L. & Tengonciang, A. M. P. Rapid appraisal of rainfall threshold and selected landslides in Baguio, Philippines. Nat. Hazards 78 , 1587–1607 (2015).

Chen, L. et al. Failure probability assessment of landslides triggered by earthquakes and rainfall: A case study in Yadong County, Tibet, China. Sci. Rep. 10 , 16531 (2020).

Article   CAS   PubMed   PubMed Central   Google Scholar  

Caine, N. The rainfall intensity—Duration control of shallow landslides and debris flows. Geogr. Ann. Ser. B 62 , 23–27 (1980).

Mandal, P. & Sarkar, S. Estimation of rainfall threshold for the early warning of shallow landslides along National Highway-10 in Darjeeling Himalayas. Nat. Hazards 105 , 2455–2480 (2021).

He, S., Wang, J. & Liu, S. Rainfall event-duration thresholds for landslide occurrences in China. Water 12 , 494 (2020).

Lin, Q., Wang, Y., Glade, T., Zhang, J. & Zhang, Y. Assessing the spatiotemporal impact of climate change on event rainfall characteristics influencing landslide occurrences based on multiple GCM projections in China. Clim. Change 162 , 761–779 (2020).

Melillo, M., Brunetti, M. T., Peruccacci, S., Gariano, S. L. & Guzzetti, F. Rainfall thresholds for the possible landslide occurrence in Sicily (Southern Italy) based on the automatic reconstruction of rainfall events. Landslides 13 , 165–172 (2016).

Hong, Y. et al. The influence of intense rainfall on the activity of large-scale crystalline schist landslides in Shikoku Island, Japan. Landslides 2 , 97–105 (2005).

Wieczorek, G. F., Morgan, B. A. & Campbell, R. H. Debris-flow hazards in the Blue Ridge of Central Virginia. Environ. Eng. Geosci. 6 , 3–23 (2000).

Kim, S. W. et al. Effect of antecedent rainfall conditions and their variations on shallow landslide-triggering rainfall thresholds in South Korea. Landslides 18 , 569–582 (2021).

Monsieurs, E., Dewitte, O. & Demoulin, A. A susceptibility-based rainfall threshold approach for landslide occurrence. Nat. Hazard. 19 , 775–789 (2019).

Dikshit, A., Sarkar, R., Pradhan, B., Acharya, S. & Dorji, K. Estimating rainfall thresholds for landslide occurrence in the Bhutan Himalayas. Water 11 , 1616 (2019).

Jordanova, G. et al. Determination of empirical rainfall thresholds for shallow landslides in slovenia using an automatic tool. Water 12 , 1449 (2020).

Article   MathSciNet   Google Scholar  

Vaz, T. et al. Regional rainfall thresholds for landslide occurrence using a centenary database. Nat. Hazard. 18 , 1037–1054 (2018).

Aleotti, P. A warning system for rainfall-induced shallow failures. Eng. Geol. 73 , 247–265 (2004).

Zhao, B. et al. Probabilistic thresholds for landslides warning by integrating soil moisture conditions with rainfall thresholds. J. Hydrol. 574 , 276–287 (2019).

Chinkulkijniwat, A., Salee, R., Horpibulsuk, S., Arulrajah, A. & Hoy, M. Landslide rainfall threshold for landslide warning in Northern Thailand. Geomat. Nat. Hazard Risk 13 , 2425–2441 (2022).

Bazhong Natural Resources and Planning Bureau. Current Situation of Geological Disasters in Bazhong City . http://zrzygh.cnbz.gov.cn/public/6596491/18173861.html (2015)

Cao, X., Wen, Z. & Chen, H. Landslide development characteristics and preventive measures in the area with red beds in Bazhong City, Sichuan Province. Chin. J. Geol. Hazard Control 30 , 20–24 (2019).

Lu, Y., Li, J. & Zhang, Q. Stability and threshold analysis of red soil landslide under rainfall conditions—Taking Erhuangping landslide in Nanjiang County as an example. Gansu Water Resour. Hydropower Technol. 51 , 15–18 (2015).

Yu, B. et al. A prediction model for rock planar slides with large displacement triggered by heavy rainfall in the Red bed area, Southwest, China. Landslides 18 , 773–783 (2021).

Zhang, Y., Wen, Z. & Cheng, Y. A discussion of the relationship between landslide disaster and rainfall types in Bazhong of Sichuan. Hydrogeol. Eng. Geol. 47 , 178–182 (2020).

Li, M. Characteristics and risk assessment of landslide hazards in Bazhou District of Bazhong City (in Chinese) (Southwest University of Science and Technology, 2016).

Läuter, H. & Silverman, B. W. Density Estimation for Statistics and Data Analysis (Chapman & Hall, 1986).

Gao, Z. et al. Comprehensive comparisons of state-of-the-art gridded precipitation estimates for hydrological applications over Southern China. Remote Sens. 12 , 3997 (2020).

Qiu, C., Ding, L., Zhang, L., Xu, J. & Ma, Z. Quantitative characteristics of the current multi-source precipitation products over Zhejiang Province, in Summer, 2019. Water 13 , 334 (2021).

Hong, T., Li, H. & Chen, M. Comprehensive evaluations on the error characteristics of the state-of-the-art gridded precipitation products over Jiangxi Province in 2019. Earth Space Sci. 8 , 1787 (2021).

Padulano, R., Rianna, G. & Santini, M. Datasets and approaches for the estimation of rainfall erosivity over Italy: A comprehensive comparison study and a new method. J. Hydrol. Reg. Stud. 34 , 100788 (2021).

Canli, E., Loigge, B. & Glade, T. Spatially distributed rainfall information and its potential for regional landslide early warning systems. Nat. Hazards 91 , S103–S127 (2018).

Cheng, L., Zang, H., Xu, Y., Wei, Z. & Sun, G. Probabilistic residential load forecasting based on micrometeorological data and customer consumption pattern. IEEE Trans. Power Syst. 36 , 3762–3775 (2021).

Crozier, M. J. Prediction of rainfall-triggered landslides: A test of the antecedent water status model. Earth Surf. Proc. Land. 24 , 825–833 (1999).

Lee, M. L., Ng, K. Y., Huang, Y. F. & Li, W. C. Rainfall-induced landslides in Hulu Kelang area, Malaysia. Nat. Hazards 70 , 353–375 (2014).

Smolikova, J. et al. Analysis of the rainfall pattern triggering the Lemesna debris flow, Javorniky Range, the Czech Republic. Nat. Hazards 106 , 2353–2379 (2021).

Lin, W. et al. Rainfall thresholds of rainfall-triggered landslides in Cili County, Hunan Province. J. Yangtze River Sci. Res. Inst. 37 , 48–54 (2020).

Khan, Y. A., Lateh, H., Baten, M. A. & Kamil, A. A. Critical antecedent rainfall conditions for shallow landslides in Chittagong City of Bangladesh. Environ. Earth Sci. 67 , 97–106 (2012).

Jiang, Z. Y. et al. Probabilistic rainfall thresholds for debris flows occurred after the Wenchuan earthquake using a Bayesian technique. Eng. Geol. 280 , 105965 (2021).

Guo, X., Cui, P., Marchi, L. & Ge, Y. Characteristics of rainfall responsible for debris flows in Wenchuan Earthquake area. Environ. Earth Sci. 76 , 596 (2017).

Melillo, M. et al. A tool for the automatic calculation of rainfall thresholds for landslide occurrence. Environ. Model. Softw. 105 , 230–243 (2018).

Huang, C.-C. Critical rainfall for typhoon-induced debris flows in the Western Foothills, Taiwan. Geomorphology 185 , 87–95 (2013).

Gariano, S. L., Petrucci, O. & Guzzetti, F. Changes in the occurrence of rainfall-induced landslides in Calabria, southern Italy, in the 20th century. Nat. Hazard. 15 , 2313–2330 (2015).

Brunetti, M. T. et al. Rainfall thresholds for the possible occurrence of landslides in Italy. Nat. Hazard. 10 , 447–458 (2010).

Martelloni, G., Segoni, S., Fanti, R. & Catani, F. Rainfall thresholds for the forecasting of landslide occurrence at regional scale. Landslides 9 , 485–495 (2012).

Wicki, A. et al. Assessing the potential of soil moisture measurements for regional landslide early warning. Landslides 17 , 1881–1896 (2020).

Valenzuela, P., Jose Dominguez-Cuesta, M., Mora Garcia, M. A. & Jimenez-Sanchez, M. Rainfall thresholds for the triggering of landslides considering previous soil moisture conditions (Asturias, NW Spain). Landslides 15 , 273–282 (2018).

Rosi, A. et al. Definition of 3D rainfall thresholds to increase operative landslide early warning system performances. Landslides 18 , 1045–1057 (2021).

Segoni, S., Rossi, G., Rosi, A. & Catani, F. Landslides triggered by rainfall: A semi-automated procedure to define consistent intensity-duration thresholds. Comput. Geosci. 63 , 123–131 (2014).

Download references

Acknowledgements

The authors sincerely appreciate the National Natural Science Foundation of China (Sino-German Mobility Programme, Grant No. M-0129) and the Science and Technology Commission of Shanghai Municipality (Grant No. 21DZ2207200) for financial support for this work. The authors thank the Natural Resources and Planning Bureau of Bazhong City for providing information concerning the landslide hazards and their locations in the period from 2011 to 2019.

Author information

Authors and affiliations.

Department of Geotechnical Engineering, College of Civil Engineering, Tongji University, Shanghai, 200092, China

Yu Huang, Cuizhu Zhao, Xiaoyan Jin, Ming Peng & Zhiyi Chen

Key Laboratory of Geotechnical and Underground Engineering of the Ministry of Education, Tongji University, Shanghai, 200092, China

Shanghai Research Center of Ocean & Shipbuilding Engineering, China Shipbuilding NDRI Engineering Co., Ltd., Shanghai, 200090, China

You can also search for this author in PubMed   Google Scholar

Contributions

Y.H.: Conceptualization, project management, funding acquisition, writing, reviewing and editing. C.Z.: Conceptualization, methodology, writing and editing. X.J.: Data Statistical Arrangement and Methodology. Y.Z.: Data processing. M.P.: Analysis of the results. Z.C.: Verification. All authors reviewed and contributed to the manuscript.

Corresponding author

Correspondence to Yu Huang .

Ethics declarations

Competing interests.

The authors declare no competing interests.

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Huang, Y., Zhao, C., Jin, X. et al. Case study of a landslide continuous probability rainfall threshold analysis based on the prediction interval principle. Sci Rep 13 , 2434 (2023). https://doi.org/10.1038/s41598-023-29625-6

Download citation

Received : 17 June 2022

Accepted : 08 February 2023

Published : 10 February 2023

DOI : https://doi.org/10.1038/s41598-023-29625-6

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

By submitting a comment you agree to abide by our Terms and Community Guidelines . If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Quick links

• Explore articles by subject
• Guide to authors
• Editorial policies

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

• Advanced Search
• Journal List
• Int J Environ Res Public Health

The Vulnerability of People to Landslides: A Case Study on the Relationship between the Casualties and Volume of Landslides in China

1 Key Laboratory of Environmental Change and Natural Disaster of Ministry of Education, Beijing Normal University, Beijing 100875, China; nc.ude.unb.liam@negiqnil (Q.L.); nc.ude.unb.liam@euxnaituil (T.L.); nc.ude.unb.liam@iqgniyuhZ (Y.Z.); nc.ude.unb.liam@iqius (Q.S.)

2 Academy of Disaster Reduction and Emergency Management, Beijing Normal University, Beijing 100875, China

Tianxue Liu

The lack of a detailed landslide inventory makes research on the vulnerability of people to landslides highly limited. In this paper, the authors collect information on the landslides that have caused casualties in China, and established the Landslides Casualties Inventory of China . 100 landslide cases from 2003 to 2012 were utilized to develop an empirical relationship between the volume of a landslide event and the casualties caused by the occurrence of the event. The error bars were used to describe the uncertainty of casualties resulting from landslides and to establish a threshold curve of casualties caused by landslides in China. The threshold curve was then applied to the landslide cases occurred in 2013 and 2014. The validation results show that the estimated casualties of the threshold curve were in good agreement with the real casualties with a small deviation. Therefore, the threshold curve can be used for estimating potential casualties and landslide vulnerability, which is meaningful for emergency rescue operations after landslides occurred and for risk assessment research.

1. Introduction

Landslides cause severe casualties and property loss in many regions. With human activities continually expanding to mountainous areas and with the global environment change, the frequency of the occurrence of landslides shows an apparent growing tendency [ 1 ]. According to the definition of Varnes [ 2 ], landslide risk refers to the expected degree of loss due to a landslide. The risk is determined by the product of the landslide hazard and the vulnerability. Landslide hazard and vulnerability information are required for a landslide risk assessment. Landslide hazard is defined as the probability that a specified magnitude of landslide occurs within a period of time and within a given area [ 2 , 3 ]. The literature concerning landslide hazard assessment is extensive [ 4 , 5 , 6 , 7 , 8 , 9 , 10 ].

However, the research on the vulnerability to landslide is relatively rare. Vulnerability refers to the degree of loss to a given element at risk due to the occurrence of a specified magnitude of hazard event [ 2 ]. Specifically, vulnerability is the relationship between the damage degree of the elements at risk and the intensity of the disaster. However, studies examining vulnerability to landslides are limited. The estimation of landslide vulnerability is highly complex due to several reasons. Glade [ 11 ] pointed out that the vulnerability of different elements at risk is variable during the similar hazard processes, and the temporal probability of people being exposed to the landslide is also variable. Moreover, the ability of different groups to cope with disasters affect the vulnerability of populations, and landslide susceptibility varies. In addition, landslide vulnerability is also affected by factors including velocity, block mass, impact angle of the landslide, position of the wall impact point, detailed geometry of the wall, and strength of the material. It is very difficult to deal with all of these factors [ 12 ].

At present, there is no uniform methodology to quantitatively assess the vulnerability of various elements at risk to different types and magnitudes of landslides [ 13 ]. Therefore, in actual landslide risk assessment, the most common method is to set the landslide vulnerability of different elements at risk to a constant 1, that is, to believe that the elements at risk will be completely damaged and lost, or to assign the vulnerability on the basis of expert knowledge and experience [ 14 ]. Several scholars study the vulnerability of building, road, and land use to landslide by the methods of vulnerability curve, vulnerability matrix, and vulnerability indicators [ 15 , 16 , 17 , 18 , 19 , 20 , 21 , 22 , 23 , 24 , 25 , 26 , 27 , 28 , 29 , 30 , 31 ]. Catani et al. [ 16 ] surveyed the loss ratio of such land uses as grassland, pasture, forest, and fruit forest under different landslide volumes and established vulnerability curves for various land uses. Remondo et al. [ 17 ] established a landslide inventory for the Deva Valley area in Northern Spain, surveyed both the relevant loss caused by landslides and the replacement costs of elements at risk in detail, and established vulnerability indicators for buildings, roads, and land in this area.

Eidsvig et al. [ 19 ] built a semiquantitative indicator-based model considering the influencing factors about a community’s ability to prepare for, deal with, and recover from the damage and loss associated with landslides to assess the relative socioeconomic vulnerability to landslides of the communities in Europe. Martha and Kumar [ 18 ] obtained the damage situation of buildings through comparing shape, color, and texture in remote sensing images before and after a landslide and assessed the vulnerability of buildings, agricultural land, and roads. Galli and Guzzetti [ 20 ] assigned the damage proportion of buildings in Umbria, Italy on the basis of several criteria and established a vulnerability threshold curve between the landslide area and the damage proportion of buildings and roads. Fotopoulou and Pitilakis [ 23 , 24 ] adopted a dynamic non-linear finite difference slope model to estimate the differential permanent landslide displacements at the building’s foundation level. The obtained differential permanent displacements are imposed at the foundation level to assess the building’s response to differing permanent seismic ground displacements based a finite element code. Finally, the fragility curves were built to assess the vulnerability of low-rise reinforced concrete buildings subjected to earthquake-induced slow-moving earth slides. Mavrouli et al. [ 25 ] used the particle finite element method to model the impact of the rockfall (velocity and size) to the masonry structure building. Then the finite element method was adopted to evaluate the mechanical properties, model the masonry structure, and calculate the internal stresses. Finally, a failure criterion was applied to calculate the damage and vulnerability of masonry structures to rockfalls were assessed. Papathoma-Köhle et al. [ 27 ] used nonlinear regression approaches to fit the vulnerability curve between the intensity of debris flow and the percentage of building loss due to the impact of debris flow. The obtained vulnerability curve was applied in Gand and Ennewasser villages in Italy. Papathoma-Köhle et al. [ 26 ] also reviewed the existing methods for physical vulnerability assessment for mountain hazards and identified difficulties in their implementation and differences between them. Finally, six future needs in the field of vulnerability assessment were proposed including user-friendliness of the method, the selection of all the relevant indicators, the transferability of the method, etc. Quan Luna et al. [ 28 ] used the FLO-2D program to model debris flow depth, impact pressure, and kinematic viscosity. Then, physical damage in terms of the ratio between monetary loss and reconstruction value was quantified. Finally, three different empirical vulnerability curves were obtained, which are functions of debris flow depth, impact pressure, and kinematic viscosity, respectively. Silva and Pereira [ 29 ] used the semi-quantitative method to assess the physical vulnerability of buildings and the associated potential losses resulting from the occurrence of shallow slides at the regional scale in the North of Portugal. Totschnig and Fuchs [ 30 , 31 ] used torrent events in the Austrian Alps to establish the quantitative vulnerability curve of the residential buildings to torrent. The results show a modified Weibull distribution and a modified Frechet distribution to fit best to the observed damage pattern in terms of absolute and relative values, respectively. Studies on examining the vulnerability of people to landslides are even more limited. Finlay [ 32 ] provided several recommended ranges and values of the vulnerability of a person in an open space, in a vehicle, or in a building derived from historical records of Hong Kong based on opinions of the expert. Murillo-García et al. [ 33 ] used the Spatial Approach to Vulnerability Assessment model to consider the indicators of Exposure, Sensitivity, and Lack of Resilience and assessed the vulnerability of population to landslide by multiplying them in every spatial unit in Pahuatlán, Mexico. Ferlisi et al. [ 34 ] assessed rockfall risk to persons traveling in vehicles along the Amalfi coastal road in Italy. The Rockfall Hazard Rating System (RHRS) method was used to qualitatively evaluate rockfall risk in different sections along the road considering the factors such as slope height, ditch effectiveness, and average vehicle risk. Then, the rockfall risk to persons traveling in vehicles was calculated under three hypothesized scenarios based on the quantitative risk assessment (QRA) procedures. The estimated societal risk is too high and cannot be tolerated. However, such models are based on the theory of disaster to build vulnerability models and has been not been subjected to enough practical testing.

Generally speaking, the non-uniform landslide intensity characteristic index and the lack of detailed historical landslide inventories have led to few studies investigating landslide vulnerability. Therefore, in this paper, the authors attempt to establish an empirical relationship between landslide intensity and the casualties caused by the occurrence of the event to fill in the gap in such studies. The authors collect and build a Landslide Casualties Inventory of China through data collection on the basis of multiple channels, and utilize the landslide cases between 2003 and 2012 as training samples to develop an empirical relationship between the volume of a landslide and the casualties caused by the occurrence of the event, that is, the casualties threshold curve, and apply this casualties threshold curve to data from 2013 to 2014 for verification. The casualties threshold curve proposed in this paper can be used to estimate possible casualties caused by a landslide quickly and provide effective information to emergency rescue.

2. Data and Methods

2.1. landslide casualties inventory of china.

China is one of the countries with the most severe geological disasters in the world. According to the statistics of China Statistic Yearbook, between 2000 and 2013, 2.5 × 10 4 geological hazards occurred every year, which resulted in the deaths of 744 persons, injured 2564 persons. Landslides and rockfalls occurred the most frequently, accounting for 90.1% of all geological hazards [ 35 ]. Figure 1 shows the distribution of landslide and rock fall disaster map of China compiled by the China Institute for Geo-Environment Monitoring [ 36 ]. Landslides and rockfalls in China are mainly distributed in the Western Sichuan Mountainous Area and the Yunnan-Guizhou Plateau Region, the Southeast Hilly Area, North of the Loess Hills, and the Qinba Mountains Area and the Northwest Plateau Region. In recent years, landslides and rockfalls occurred frequently in these areas resulting in severe casualties. Figure 2 shows four cases of landslides in Guangxi, Yunnan, and Shanxi Province. Therefore, it is urgent to have landslide risk assessments for the identification and management of potential high landslide risk areas.

Landslide and rockfall distribution map in China.

Cases of landslide: ( a ) Landslide in Baqiao District, Shaanxi on 17 September 2011; ( b ) Landslide in Quanzhou County, Guangxi on 9 May 2011; ( c ) Rockfall in Zhongyang County, Shanxi on 16 November 2009; ( d ) Landslide in Yiliang County, Yunnan on 4 October 2012.

In the present study, the term landslide refers to the wider definition of the landslide with the exception of debris flow. The reason we excluded debris flow is that the amount of casualties caused by debris flow in China is more seriously than that of other landslides. For instance, the Zhouqu debris flow occurred on 8 August 2010 in Gansu Province, northwestern of China resulted in 1765 deaths. On 10 June 2005 in Heilongjiang Province, northeast China, the shalan debris flow caused 117 deaths.

The inventory of landslides in China has collected landslide information from four different data sources. The first source is the National Geological Hazard Bulletin [ 37 ], prepared by the Geological Environment Monitoring Institute of China. The Bulletin primarily includes an overall introduction to the frequency of occurrence of geological hazards in China and the number of losses in the year, a distribution and comparison with the past, and detailed survey information of serious geological hazard. The available time range for acquiring the data of the National Geological Disasters Bulletin is between 2004 and 2014, and the annual bulletin can be downloaded from the website ( http://www.cigem.gov.cn/ ). The Geological Environment Monitoring Institute of China conducts relevant survey and research, such as geological hazard monitoring, forecasting, and early warning. The institute’s data quality is reliable; in particular, the field investigations of serious geological hazard events are accurate and credible.

The second data source is the Yesterday Disaster Report [ 38 ], prepared by the National Disaster Reduction Center of China. Since 2004, reports on disasters that have occurred in China (including earthquake, geological disaster, flood, drought, typhoon, and lightning strike) are summarized and compiled every day. The Ministry of Civil Affairs is responsible for natural disaster rescue in China; therefore, the Yesterday Disaster Report primarily investigates casualties and property loss caused by disasters and examines loss caused by disasters in detail. Landslide information that caused casualties in China is screened and obtained out of the Yesterday Disaster Report.

The third data source is the Geological Disaster/Hazard Report [ 39 ]. Since July 2011, the Ministry of Land and Resources issued an introduction on its website ( http://www.mlr.gov.cn/dzhj/dzzh/zqxqbg/index.htm ) to the geological disaster/hazard that have occurred all over China every day. The fourth data source is any news report on landslides in China, acquired through searching “geological hazard”, “geological disaster”, “landslide”, “casualties”, and other keywords on Google, to obtain rockfall and landslide information in news reports.

Finally, the authors summarize landslides that caused casualties from 2003 to 2014, collected and arranged data out of the four different data sources including the National Disaster Reduction Center, the Ministry of Civil Affairs, the Ministry of Land and Resources, the China Geological Environmental Information Website, and news searches. Landslide events that caused casualties were proofread according to the time and location of occurrence. Repeated disasters were eliminated. These landslide cases were uniformly arranged by such factors as type, occurrence time, location, intensity, casualties, and loss. At last, the Landslide Casualties Inventory of China were established including 602 landslide casualty events, 4157 casualties, and 2405 deaths. Details of any specific data source are referred in Table 1 .

Data source of landslide casualties database of China.

Based on the Landslide Casualties Inventory of China, the authors select landslide event with detailed magnitude information which can be used to develop an empirical relationship between landslide magnitude and casualties caused by the occurrence of the event. One hundred and fifty-six landslide events include detailed volume and casualty information. The authors use 100 landslide cases occurred between 2003 and 2012 to establish a landslide casualties threshold curve and apply the casualties threshold curve to the remaining 56 landslide cases between 2013 and 2014 for validation.

2.2. Methods

In this study, we used the average volume and causalities in different landslide groups to analyze and develop an empirical relationship between the volume of landslide and casualties caused by the occurrence of the event. Detailed steps are described as follows.

Among 100 landslide casualties events that occurred between 2003 and 2012 used for building the empirical curve, a landslide with the volume of 1 m 3 in Qidaohe, Xintian Town, Wanzhou District, Chongqing—due to continuous heavy rain on 21 June 2007—had the smallest volume; while a large-scale mountain landslide in Jiwei Mountain, Tiekuang, Wulong County, Chongqing on 5 June 2009 and a landslide in Baofeng Village, Guanmen, Nanjiang County, Bazhong, Sichuan on 18 September 2011 had the largest volume, 5 × 10 6 m 3 . The volumes in different landslide cases are notably different. It is unreasonable and inadvisable to analyze them all together. To solve this problem, the method of analyzing the average vulnerability suggested in the literature [ 12 ] was adopted. Therefore, according to criteria of similar volume within a group, and large volume differences between groups, the authors classified 100 landslide cases into 16 groups in accordance to their volume. Then, the average casualties caused by a landslide were calculated for each group. The representative volume and the average causalities in each landslide group can be analyzed. A detailed classification threshold, average volume, and volume standard deviation can be seen in Table 2 .

Result of landslide groups according to their volume.

Although most of the causes of casualties can be explained by the volume of landslides, there still exist some uncertainties due to the influence of other factors such as the velocity of the landslide, the location of the people (outdoors, in the building, in a vehicle), the population density and so on. To describe the uncertainties of casualties caused by landslides, the authors use an error bar and its standard error. The formula is as follows:

where SE is standard error, S is sample standard deviation, x is sample value, x ¯ is sample average value, and n is sample number.

The standard error of casualties caused by landslides in groups of different volumes was calculated according to the Formula (1). The upper limit and lower limit threshold value of casualties caused by the occurrence of the landslide event were obtained by plus/minus one standard error on the basis of the average value. An error bar was adopted to indicate the upper limit and lower limit threshold value of casualties, that is the uncertainty.

The final step is to fit the upper limit or lower limit threshold value curve of casualties caused by a landslide. In this study, a linear function, exponential function, logarithmic function, high order polynomial, and power function were used to fit the threshold curve. The R Square was applied to assess the fitting results.

3. Results and Validation

For average casualties caused by a landslide in groups of different volumes, see Figure 3 . The horizontal axis shows the logarithmic volume of a landslide, while the longitudinal axis shows the average casualties caused by that landslide. The figure shows that the casualties caused by the occurrence of the landslide are obviously related to the volume of the landslides. With the gradual increase of the volume of a landslide, the casualties caused by the occurrence of the event also increase constantly. For example, when the volume of a landslide is less than 1000 m 3 , the average casualties are within 10, when the volume of the landslide is 1000 m 3 –500 × 10 3 m 3 , the casualties are approximately 10–20, and when the volume of a landslide is more than 500 × 10 3 m 3 , more casualties are caused. There are still casualties caused when there is nearly 0 volume of landslide in Figure 3 . It is because these landslides occurred along the road or the scenic areas which increased the probability of casualties. The size of the error bar also shows that when the volume of a landslide is small, the uncertainty of casualties caused is small. With the increase of the volume of a landslide, the uncertainty exhibits a growing tendency.

Empirical relationship between volume of landslide and the casualties caused by the occurrence of the event (the black dot in the figure indicates average casualties, the error bar is standard error, indicating the uncertainty of casualties caused by landslide; thin line/thick line respectively indicate the maximum and minimum threshold values of the landslide casualties curve).

The fitting results of different functions are shown in Table 3 . For the maximum threshold curve, the power function has the highest R-square with the value of 0.973. For the minimum threshold curve, the logarithmic function has the highest R-square with the value of 0.974 following by the power function with an R-square of 0.964. Based on the fitting performance of the maximum/minimum threshold curves, the power function was finally selected to represent the threshold value curve of casualties caused by landslides. In addition, it has been shown in the literature that the power function is suitable for the fitting of the landslide vulnerability curve. Galli and Guzzetti [ 20 ] adopted the power function to fit the landslide vulnerability curve because the power function is simple and performs reasonably well. Furthermore, the power function suggests a self-similar scaling behavior of the landslide damage which allows the damage information for small landslides to be applied for estimation of the vulnerability to large landslides and vice versa. The fitting result of the power function, that is the maximum and minimum landslide casualties threshold curves, are shown in Table 4 .

Model fitting summary.

Maximum/Minimum threshold value curve of casualties caused by landslide.

The landslide casualties threshold curve that was established based the landslide cases occurred between 2003 and 2012 was applied to 56 landslide cases which were occurred between 2013 and 2014 to verify the results of the landslide casualties threshold value curve. In 56 landslide cases from 2013 to 2014 with detailed intensity and casualties, a rockfall in Huangchang Village, Mawu Town, Shizhu Tujia Autonomous County, Chongqing on 10 September 2013 had the smallest rockfall volume—2 m³—while a landslide in Changli County, Qinhuangdao, Hebei on 16 November 2013 had the largest rockfall and landslide volume—4 × 10 6 m³. The volume of the landslides are within 1–5 × 10 6 m³ and can be used to validate the landslide casualties threshold curve.

Table 5 shows the information of landslide casualty events with a volume of more than 1000 m³ among the landslide cases from 2013 to 2014, and the actual casualties are compared with the casualties estimated through the landslide casualties threshold curve. In 21 of 26 landslide events, casualties are correctly estimated with an accuracy rate of more than 80%. In the landslide events which the casualties are wrongly estimated by the landslide casualties threshold value curve, the estimate deviation is used to assess the estimated error. Table 5 shows that the deviation of the wrong estimation is small. Except for landslide events No 17, 21, 24–26 the casualties estimate deviations in the rest of the landslide events are correct. For event 24—the largest error—a landslide in Qingchuan County, Sichuan on 12 July 2013 ( Figure 4 a) was 20 × 10 4 m³, 19 houses were damaged, but only one casualty was caused. When the landslide occurred at 16:10 p.m., adults were working while children were studying at school; therefore, casualties were few. In event 17, a 150 × 10 4 m³ landslide ( Figure 4 b) in Shengli Village, Gukai, Nayong County, Guizhou on 16 May 2013 was caused by the demolition of an aggregate plant. There were three casualties, including two demolition employees of the NaYong Demolition Company and an employee of the plant, while villagers nearby had been transferred to a safe area, which prevented serious casualties. In events 21, 25 and 26, the error is more than 50%. The actual casualties in the events are 1–2, and it is difficult to forecast such a small casualty event precisely. However, for a small casualty event, a 50% forecasting error will not have a strong influence on the level of emergency rescue measures.

( a ) Landslide in Qingchuan County, Sichuan; ( b ) Landslide in Nayong County, Guizhou.

Validation of the threshold value curve of landslide casualties.

1 Deviation = Min (estimate threshold value-actual value)/actual value).

Therefore, the results of Table 5 show that the landslide casualties threshold curve established in this paper is good and can be applied in estimating casualties caused by landslides.

4. Discussion

By compiling a Landslide Casualties Inventory of China, the authors attempt to develop empirical dependencies between the volume of landslide and casualties caused by the occurrence of the event. From the average status of casualties caused by a landslide, when the volume of a landslide is less than 1000 m 3 , the average casualties are within 10; when the volume of a landslide is 1000 m 3 –500 × 10 3 m 3 , the casualties are approximately 10–20; and when the volume of a landslide is more than 500 × 10 3 m 3 , more casualties are caused. In addition, due to complexity of factors that affect casualties caused by a landslide, the specific casualties caused by a landslide with a certain volume will exhibit fluctuations around the average value. The size of the error bar in Figure 3 shows such uncertainty. The fluctuation of the error bar indicates that the increase in the volume of a landslide results in increasing uncertainty of casualties. The validation results suggested that the proposed method considering the average status and uncertainty for establishing an empirical relationship between landslide intensity and the loss caused by that landslide can be further applied to other geological hazards or a smaller scale with a complete landslide intensity and loss inventory.

However, casualties caused by landslides are affected by many factors, including velocity of landslide, specific type of landslide, location, time of occurrence and population density, the location of the people (outdoors, in the building, in a vehicle), plus any hazard defenses and resisting capacity (e.g., age and gender). Finlay [ 32 ] analyzed the influence of the location of the person (in an open space, in a vehicle, or in a building) on the vulnerability of a person based on experts’ knowledge. García et al. [ 33 ] used the indicators—including the total population index, population density index, population lower than 12-year-old index, female population index, etc.—to assess the vulnerability of population to landslide in every spatial unit in Mexico. Therefore, estimating casualties caused by landslides is a complex task. The landslide casualties threshold curve obtained in this paper from the development of an empirical statistical relationship between the volume of a landslide and the casualties caused by the occurrence of the event is a useful attempt, but there are still some limitations in actual application. For example, the landslide casualties threshold value curve is established on the basis of a landslide; therefore, it cannot be applied in assessing vulnerability to a debris flow and the casualties or loss caused by a debris flow. The indicator of landslide intensity also only considers the volume of a landslide; because the maximum landslide volume in Landslide Casualties Inventory of China is only 5 × 10 6 m³, it cannot be applied to assessing landslides with a larger volume.

Two factors lead to limitation above: one is the incomplete landslide inventory. The scattered occurrence of landslides makes it difficult to collect data and arrange a large number of complete landslide casualties and loss information. Therefore, in this paper, the authors can only utilize limited landslide cases with detailed information, specifically 100 landslide events occurred between 2003 and 2012 to establish the maximum/minimum casualties threshold curve and 56 landslide events between 2013 and 2014 were used to verify the landslide casualties threshold value curve. The casualties caused by the occurrence of landslides are affected by many factors, which means that a landslide with the same volume may result in completely different casualty situations. Only a part of the casualties caused by a landslide can be explained through the volume of the landslide; there is a certain error that is reflected by the error bar results in Figure 3 . Therefore, if the landslide casualties threshold curve can be established on the basis of a complete landslide database, a more stable and reliable result will be obtained, which should be determined and discussed in future studies.

5. Conclusions

In this paper, the authors collected information regarding casualties caused by the occurrence of the landslides in China from 2003 to 2014 and established the Landslide Casualties Inventory of China through multiple information sources. On the basis of landslide events that occurred between 2003 and 2012, the authors developed an empirical relationship between volume of landslides and casualties caused by the occurrence of the event. Utilizing an error bar to describe the uncertainty of casualties caused by a landslide, a landslide casualties threshold value curve is established. The authors verify the curve by using landslide events that occurred between 2013 and 2014, and the results indicated that this threshold value curve has good agreement with estimated casualties.

The landslide casualties threshold value curve can be used to estimate possible casualties caused by a landslide quickly and provide effective information to emergency rescue. Conversely, this curve can also be used to identify the vulnerability of people to landslides to a certain extent and can provide information for research on a quantitative landslide risk assessment study by combining the curve with landslide hazard assessment models.

Acknowledgments

This work was supported primarily by the National Key Research and Development Program of China “Study on risk assessment model for Population and economic system under Global change” (2016YFA0602403), National Key Technology R&D Program of 12th Five Year Plan of China (2012BAK10B03), and the National Natural Science Funds (41271544).

Author Contributions

Ying Wang designed the study and revised and approved the final manuscript, Qigen Lin collected the data, wrote the first manuscript, and revised and approved the final manuscript; Tianxue Liu, Yingqi Zhu, and Qi Sui collected the data, conducted data analysis, and revised and approved the final manuscript draft.

Conflicts of Interest

The authors declare no conflict of interest.