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

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.

1. 1.

   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.

2. 2.

   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.