Remote monitoring techniques for rehabilitated slope

AGL Loy Yang has been intensively engaged and involved in research in the area of rehabilitated mining slopes. A proposal was raised to turn the Loy Yang mine pit into a lake after decommission (in three decades) to achieve safe and stable rehabilitation of the mines. The slope stability of these rehabilitated slopes is one of the major concerns. The slopes are hardly failed spontaneously. Instead, they usually provide indications of distress (cracks and erosion) over some time. Therefore, an ongoing monitoring system may provide valuable time to mitigate the progression of the failure. Distributed optical fibre sensing (DOFS) and unmanned aerial vehicle (UAV) are emerging new and innovative technologies for remote monitoring a large civil structure for early warnings, alerts and decision making. Both DOFS and UAV have been recently deployed at Loy Yang rehabilitated trial site. The preliminary results have demonstrated their reliability and practicality for these rehabilitated slope monitoring techniques. These remote sensing techniques are compared and discussed in terms of slope monitoring application.


Introduction
AGL Loy Yang's open-cut coal mine is located 165 km southeast of Melbourne, Australia. AGL Loy Yang Power Station provides 50% of Victoria's energy requirements [1]. However, the AGL's mining licence will expire in another three decades, and the mine is required to undergo progressive rehabilitation. Rehabilitation of the site involves cutting and filling benches of the inactive mine and capping with layers of clay, fill material, and topsoil to ensure the stability of the rehabilitated slopes.
Distributed optical fibre sensors (DOFS) and unmanned aerial vehicles (UAV) have both gained worldwide attention and have successfully demonstrated their potentials for structural health monitoring (SHM) applications [2,3]. Both techniques are well known to be able to monitor large-scale engineering civil structures. DOFS can monitor localised temperature, strain and vibration along a single strand of fibre cable, whereas optical camera integrated UAV can take aerial images and reconstruct 3D models using photogrammetry. However, both remote monitoring techniques are passive sensing, which requires continuous monitoring to provide valuable data for SHM purposes.
Matano et al. [4] reported implementing an integrated fibre optic system to control the rock slope stability in the Coroglio tuff cliff at the border of the active volcanic caldera of Campi Flegrei. Schenato et al. [5] used a distributed optical fibre sensing system to measure landslide-induced strains on an optical fibre buried in a large-scale physical model of a slope. Lanciano and Salvini [6] reported the monitoring activity successfully tested in a marble quarry by Brillouin sensing combined with drone photogrammetry and geotechnical survey. These monitoring techniques have significantly improved the safety at work as well as assisted the extraction planning. In addition, the work has demonstrated the effective capability of DOFS techniques to monitor complex geomorphological sites.
There are still multiple research components that need further investigations before implementing UAV and DOFS for slope monitoring as they are both considered relatively new. By corporating with AGL Loy Yang, both techniques have been explored and recently deployed at one of the rehabilitated trial sites to investigate both remote monitoring techniques' performance. The paper will present some preliminary results obtained from both methods and make some comparisons between them.

Methodology
AGL Loy Yang provides one of the rehabilitated trial sites, see Fig 1(a), to explore different monitoring techniques at the open cut coal mine. The site was cut and filled to provide a gradient of 3:1 (H: V), see Fig 1(b). The slope was then covered with a thick layer (varying from 0.5 m to 2 m depth) of clay and 0.1 m of topsoil.
For the study, military graded optical fibre cable sourced from FS.com [7] was proposed to be directly buried in the clay layer at a depth of 0.5 m from the surface for subsurface monitoring. Several layout plans were drafted for installing the fibre cable across the trial slope. After considering the possible overall failure of the rehabilitated soil cover and potential local failures on the slope surfaces, it was decided to lay two fibre optic cables as shown in Fig. 1 (c). The total length of the fibre installed on this slope was 1.3 km and 300 m for cable 1 and cable 2, respectively. A contractor was hired to trench a 0.5 m deep trench line on the trial site, and the optical fibre was laid down at the bottom of the trench, see Fig  2. The optical fibre cables are buried inside the clay capping with the intention to (1) reliably capture any displacement of the clay capping and (2) avoid interference due to the roots of vegetation. After installing the optical fibre cable, the trench line was backfilled and lightly compacted using a hand compactor. The location of the installed fibre cable was tracked with geographic information system (GIS) mapping. For the field trial, two measurements (six months apart) were taken using Brillouin optical time-domain analyzer (BOTDA) based Distributed Strain Temperature Sensor (DSTS) from OzOptics. The duration of each measurement is approximately 1 minute. The specifications of the BOTDA is listed in Table  1.
UAV flight was piloted by a third party licenced remote pilot aircraft system (RPAS) to take the aerial images over the entire western rehabilitated site (140 hectares or 1.4 km 2 ). Due to the limit of battery life of the UAV and different elevation levels, the route was split into eight different paths to cover the scan of the entire western rehabilitated site. Each flight path was set up using a commercialised software, Pix4DCaputre [9]. The specifications of the UAV used for the scanning is also listed in Table 1. The total time taken to scan the region of interest is approximately 2 hrs. Due to the current budget, the UAV scan can only be conducted once every year. In addition, multiple 1.6 m long EnviroPro soil probes and multiple level sub-surface instruments sourced from EnviroPro® [8] were installed across the entire rehabilitated slope. EnviroPro soil probe is a commercialised product in Australia that has recently gained attention to the utilities. The probe is designed to measure temperature and volumetric soil moisture at multiple points along with the probe. The longest probe (1.6 m) used can measure both temperature and soil moisture with a spatial resolution of 0.1 m. It is self-powered using a solar panel, and the data can be monitored and uploaded to the cloud every 30 minutes. The cost for each EnviroPro soil probe is A$2000-4000 (depending on the specification and length). The soil probe is discrete sensing, and each probe can only monitor the information around a very localised area of 0.3 m × 0.3 m. The calibration is necessary to provide accurate and precise insight into the capping and hence justify the condition of the capping for the rehabilitated slope. Furthermore, the soil probe will also measure temperature information at a depth of 0.5 m where the fibre cable was installed for temperature compensation purposes. The detail of EnviroPro is listed in Table 1.

Distributed optical fibre sensing (DOFS)
Brillouin frequencies along the two of the buried cables were recorded using the DSTS unit over time (six months apart) and presented in Fig 3(a) and (b), respectively. The Brillouin frequencies shift, Δvb is known to have a linear relationship to the change in both strain and temperature [12,13]. From the temperature measurements obtained from EnviroPro at the trial site, the two scans' temperatures at 0.5 m depth were 12.11 o C and 14.38 o C, respectively. The strain, ɛ, can then be calculated using Eqn. 1.
where ɛ f is the fibre strain measured along fibre direction, C ɛ and C T are the strain and temperature coefficients respectively, Δvb is the Brillouin frequency shift, also known as the difference between the latest Brillouin frequency and the reference Brillouin frequency, vb(t1) -vb(t0), and ΔT is the temperature difference. The calculated strain with temperature compensated is also plotted in Fig 3. In Fig 3a, a few spikes can be observed around 900 m from the starting point along with Cable 1. Once the cable is buried, the actual location of the fibre cable will be hard to identify. It is assumed that the fibre cable will remain within the buried area with a tolerance of ±0.5m (in easting and northing direction). By combining the GIS information of the buried cable, the fibre strain can be presented in a 2.5-D space, as shown in Fig 4(a), with the colour contour as  . Fig 4(b) shows a grid data interpolation of the strain information obtained from cable 1 using MATLAB to identify the location of anomalies. The highest strained region is noted to be close to the edge of the rehabilitated trial site (top left corner), where other rehabilitation work took place. As indicated in Fig 4b, the spikes can be located, and it is noted to be at the corner of the constructed slope (top left). The high strain measurement can be due to the localised ground movement due to the newly constructed slope next to the trial site.
Otherwise, the strain readings can identify no alarming movements apart from extremely localised movements caused by shrinking and swelling. This was also confirmed by visual site inspection, where no large cracks or slides were observed. Strain fluctuations in both Fig  3a and 3b correspond to these minor movements within the clay capping. At the current state, there is no clear strain threshold for providing the alert or warning at this stage. Hence, more work will be needed to determine the static or dynamic threshold to set up the warning alert.   Type of optical fibre selection is also crucial for the monitoring work. Military graded optical fibre sensor is chosen due to its robustness against crushing and pulling. The cost for the cable itself is approximately A$20-30 per metre. However, the long-term performance of slope monitoring is still yet to be investigated. A broken cable can be repaired and re-spliced, given that the location of the buried fibre can be tracked. However, it can be hard to repair once the cable is damaged due to the site's condition (buried under 0.5 m from the surface), requiring more trenching and potentially damaging other sensors.
The direct burial method is trialled at the rehabilitated site as it requires less technical skills by just laying the optical cable down the trench line and avoiding kinking. However, this installation method will have low cohesion between the fibre and soil, which will affect the accuracy of the displacement measurements (calculated from the measured strain). Furthermore, the direct burial method requires trenching and backfilling process, which increases the cost for installation. Erosion can potentially happen along the trench line as that is where the water can easily find its way down the slope.
One of the biggest challenges in using fibre optic sensors for geotechnical engineering applications is to ensure the adhesion between soil and fibre optic cable. When the adhesion between the fibre and soil is weak, fibre will slip through the soil, causing a relative movement that hiders the measurement of actual soil movement. In the field trial presented in this paper, the main objective was to test the ability of DOFS fibre to capture large soil movements, which could trigger slope failure. Thus, any possible slippage was ignored. Results presented in Fig. 3 and 4 shows that fibre has captured local soil movements along the slope. This implies that DOFS can be successfully applied to monitor ground movement in large geotechnical applications despite the possibility of relative movement. However, it must also be noted that relative movement (slippage) can bring considerable errors to results in applications where precise soil displacement/strain is to be measured. Therefore, quantification of slippage against soil type and condition (moisture content) is needed to correct the measured data in these applications.

Unmanned Aerial Vehicle (UAV)
A total of 1277 aerial images were captured over the western rehabilitated site. Metashape Professional by Agisoft [14] was used to perform photogrammetry analysis. All the geotagged aerial images are first loaded, aligned and stitched. High configuration dense cloud is constructed with more than 9 aerial images overlapping in most regions, see Fig 5(a). The black dots in Fig 5(a) indicated the location of each aerial image. A digital elevation model (DEM) is constructed with 12.2 cm per pixel spatial resolution, see Fig 5(b). A section of the 3D model and geo-tagged aerial images around the trial site is presented in Fig 6. UAV is an attractive and emerging remote sensing technology. The aerial images can be used to provide global and scenic views for presentation purposes and construct a 3D model, digital surface model (DSM), and digital elevation model (DEM) for further finite element analysis. However, constructing an accurate model of such a large scale (14 hectares) will require lots of ground control points (GCPs) across the site. For the trial, only two GCPs were used by the third-party RPAS. Hence, the accuracy of the model created is yet to be validated. It is also noted that the vegetation on the slope can affect the interpretation of the slope profile when reconstructing the 3D model. Any crack and anomaly can be partially or fully obscured by the vegetation, causing some interpretation error.  Although most of the work is autopiloted using Pix4DCapture by the RPAS, it will still require a proper RPA licence by CASA, adequate planning according to the operating licence and insurance to fly the UAV in Australia. This increases the cost for performing the scanning job (A$8000-$10,000 per scan). The scanning can be controlled remotely, but the monitoring work will require the pilot to go to the site to set up the flight physically. Each flight will take 40 minutes (max) and require replacing batteries. To scan the entire western rehabilitated site, it will take up to 1-2 days. Due to the budget, the scans can only be done on a yearly basis.
Nevertheless, the area coverage of UAV is significantly larger than the existing monitoring technique. In addition, this method does not require any intrusion to the slope, which is preferable by the mining operation and maintenance team. However, for surface displacement monitoring, it will require to have some traceable features on the slope to track the displacement. At this stage, the tracking process heavily relies on manually identifying and tracing the existing features like power lines, weather stations, survey pins and aboveground pipelines. Therefore, more work will be needed to be done regarding placing some artificial features or markers on the slope and using computational vision to improve the tracking process over time.

Moisture monitoring
Since the soil properties can vary across the site, the soil sample at the instrumented site was bored and brought to the laboratory. The volumetric water contents for topsoil, clay and coal were calibrated separately under laboratory conditions. The calibrated volumetric water contents and temperature data obtained at four different depths from one of the EnviroPro instrumented across the rehabilitated slope are presented in Fig 7(a) and 7(b). The rainfall and evaporation data obtained from the weather station close to the trial site are reported in Fig 7(c).
Water content measurements obtained from probes are useful to estimate the pore pressure rise due to precipitation. Loy Yang mine is situated in Latrobe Valley, which has a wet temperate climate. Rainfall-induced shallow slope failures occur during short and intense precipitations or after long rainy periods, depending on the infiltration capacity and hydromechanical properties of the involved soil. Typically, these slope failures are characterised by a thickness ranging from some centimetres to a few meters (usually 1-2 m) and a length to thickness ratio greater than 10. In addition, shrinkage cracks on the surface facilitate the infiltration and promote shallow slopes failures exposing the coal layer to the atmosphere. According to the volumetric water content measurements shown in Fig. 7(a), there was no significant rise in pore water pressure throughout the trial period. In addition, the rainfall and evaporation data agrees well with the volumetric water contents measured using the EnviroPro. However, to accurately calibrate the soil moisture readings, it will require a soil sample at every installation site, which can be time-consuming and costly. Nevertheless, the EnviroPro can qualitatively provide the moisture content of the capping and hence the performance of the capping.

Conclusion
This paper explored the deployment of different emerging remote sensing techniques for rehabilitated slopes, including distributed optical fibres sensing, UAV photogrammetry and EnviroPro soil probe at AGL Loy Yang rehabilitated trial site in Victoria, Australia. Each of the techniques can provide valuable information for the operation team to better monitor the rehabilitated slope. However, it is also noted that none of the monitoring techniques is perfect for rehabilitated slope monitoring work. Therefore, more future work, especially in the area of artificial intelligence, computational vision, data analytics and data fusion, will enhance the performance of these techniques. Furthermore, the long-term performance, repairability and robustness of each method are yet to be investigated. The following summarised the findings and lessons learnt from these remote sensing technologies for subsurface rehabilitated slope monitoring.  Two strands of optical fibre cables (1.3 km and 290 m) were buried directly along the rehabilitated slope at a depth of 0.5 m from the surface. This installation method required trenching, backfilling and compaction, which can be costly and time- consuming. In addition, erosion can happen along the trench line since the water can find an easier path down the slope.  BOTDA-based DOFS was used to measure the strain along the optical fibre cable with a spatial resolution of 1 m. The temperature at 0.5 m depth obtained from the EnviroProb soil probe was used to compensate for temperature change. Once the cable is buried, the actual location of the cable will be hard to identify. By assuming the cables remain within the location after being buried, GIS information of the cable (before burial) can be used to present the strain information in 3D space.  The strain measurement obtained from DOFS shows that the direct-buried cables can identify local soil movements along the slope. However, slippage can occur when the adhesion between the soil and cable is weak and affects the accuracy of the strain measurement of DOFS. Therefore, quantification of slippage against soil type and condition (moisture content) is needed to correct the measured data in these applications.  UAV is an attractive and emerging remote sensing technology. A total of 1277 aerial images are used to construct a 14 hectares 3D model, which can be exported for further finite element analysis. However, constructing an accurate model of such a large scale (14 hectares) will require lots of ground control points (GCPs) across the site. Furthermore, each scan can cost AUD 8,000 to 10,000, and hence, it can only be done on a yearly basis due to the current budget.  The area coverage of UAV is significantly larger than the other existing monitoring techniques. In addition, this method does not require any intrusion to the slope, which is preferable by the mining operation and maintenance team. However, surface displacement monitoring will require some traceable features on the slope to track the displacement. At this stage, the tracking process heavily relies on manually identifying and tracing the existing features like power lines, weather stations, survey pins and above-ground pipelines. Therefore, more work will be needed to be done regarding placing some artificial features or markers on the slope and using computational vision to improve the tracking process over time.  The vegetations on the slope can affect the profile of the slope when reconstructing the 3D model. In addition, any crack or anomaly on the slope may also be fully or partially obscured by the vegetation, causing interpretation error.  EnviroPro soil probe is a commercialised product that has been designed to provide localised volumetric water content and temperature at different depths with a spatial resolution of 0.1 m. It is self-powered using a solar panel, and the data can be monitored and uploaded to the cloud every 30 minutes.  As the soil probe is instrumented through multi-layer capping, the calibration is necessary to provide accurate and precise insight into the capping and hence justify the condition of the capping for the rehabilitated slope.
We would like to thank AGL Loy Yang for the necessary funding and data for this study.