Catalogue of real-time instrumentation and monitoring techniques for tailings dams

With ongoing catastrophic mine tailings dam failures, the hindsight revelation of poor safety records, and an increasing prevalence of public scrutiny and attention of mining operations, there is an immediate call for enhanced safety of tailings dams. Today, challenges arise in identifying and utilising the ability of monitoring systems to understand the complex performance and behaviours of these dams, reflecting on the system's ability to predict deterioration before failure occurs. New literature, mining regulators, insurance firms, and mining practitioners are calling for increased diligence in the form of real-time monitoring: but what can the industry offer in response? This research establishes a centralisation of real-time monitoring instruments suitable for tailings dams, discussing the specifications, advantages, and disadvantages of each. An understanding of traditional instrumentation, online monitoring systems, and the value of centralised monitoring was detailed. Collaboration with suppliers discovered innovative systems which enable monitoring of different failure modes and mechanisms.


Introduction
Real-time monitoring provides the opportunity to have continuous readings of different instrumentation that are installed within the tailings dam structure, automatically presented in an interpretable format and notifying key personnel of issues before they have even considered scheduling a time to analyse the data.This is an advancement on traditional approaches, whereby in order to collect and interpret the data, mine personnel are required to: (1) Visit the instrumentation site; (2) Download the data; (3) Return to the office; (4) Transfer/input the data into the system; (5) Manipulate the data into an interpretable form; (6) Interpret the data (or communicate the data to the interpreting party, which may then take in the order of weeks to months to receive feedback); then (7) Respond as appropriate.
There are a number of inherent risks associated with this process.The three most significant risks are the time between readings, potential for human error, and hazardous access.The time between collection and reporting can vary from anywhere between a day to a month, giving rise to the question of how truly representative the data is of dam performance by the time it is analysed.This process is also challenged by the speed at which deterioration of tailings dams progresses to failure; the monitoring process is near redundant if it is not appropriately frequent to catch deterioration and respond while the opportunity exists.Human error with data collection, input, transfer, manipulation also exists: human error can be caused by time pressures, complacency, or for the simple fact that a mistake was made.Changes in personnel can also complicate data handling and retention, where familiarity and undocumented knowledge can easily be lost.By automating the data handling, this risk is mitigated and also allows mine personnel to reallocate this time to critical interpretation of the data.Hazardous access is particularly prevalent during times of deterioration.Should an area be deemed unstable, it is important to understand its behaviour but not at the risk of sending a human to collect a measurement.Real-time, remote monitoring mitigates this exposure.
Real-time instrumentation works to counter these (and other) limitations by (Clarkson et al. 2020): . Automating data handling to reduce the risk of human error and allowing mine personnel to reallocate the time to interpretation of the data; .Centralising monitoring data that can be queried by authorised users at any time, from anywhere in the world; . Exhibiting increased monitoring programme reliability while reducing data acquisition and processing costs; and .Allowing current and historical data interpretation with interaction of any data along the timescale.
Literature has documented a considerable overlap between the instrumentation recommended for embankment dams when compared to concrete dams (Avella 1993).However, the embedment of real-time capabilities appears to be on a case-by-case basis (aside from seismic monitoring), and as such is not readily documented.Comparing the mandated read frequency between embankment dams (Clarkson and Williams 2019) and concrete dams (Avella 1993), the following key observations are noted: . The minimum recommended inspection frequency for seepage measurements and pore pressures is the same; .For piezometers, observations wells, foundation deformation, and extensometers, concrete dam readings were recommended at a frequency of two to six times higher than embankment dams; and .For geodetic surveys (EDM, theodolites, etc.), embankment settlement points, and total pressure cells, concrete dam readings were recommended at a frequency of 10-12 times higher than embankment dams.
ICOLD (2001) reported that in the 18,000 mines around the world, the failure rate of tailings dams in the past 100 years was estimated at 1.2%, while the failure rate of the traditional water storage dam was 0.01%.
There has been speculation around recent, catastrophic tailings dam failures, that the monitoring would not have been able to predict or foresee the failure.Yet, the full opportunity presented by instrumentation and monitoring systems has not yet been realised or implemented.Until the global practice of tailings dams monitoring is improved, and opportunities are utilised in practice (hence proving whether or not these systems are able to predict tailings dam failure), it is inappropriate to deem this an insufficient method for identifying failure ahead of time.

Method
This paper forms one part of a wider research project by the authors to develop a comprehensive monitoring strategy for tailings dams.This paper aims to identify and catalogue the present-day range of tailings dam monitoring instrumentation and technologies.Supplementary to this is a separate paper (Clarkson and Williams 2020) which describes the systems and networks required to implement the instrumentation and technologies described.
To complement the research and enhance applicability in the industry context, a number of suppliers were engaged with 24 suppliers providing input to this research paper.A scrutinised compilation of the feedback provided is presented within this paper.The contributing suppliers are acknowledged on Page 9 of Appendix A of this paper.
While static liquefaction has become a topic of interest in recent years, the focus in mitigating the risk associated with this has been on design considerations of potential strength reductions, operational control and planning, and material characterisation for susceptibility to liquefaction and to understand pre-consolidation stresses.KCB (2018) stated that 'much of the risk depends on the in-situ stress regime, which is difficult to measure and monitor'.Considering the speed of static liquefaction occurrence, and the current inability of monitoring systems to identify conditions preceding static liquefaction, real-time monitoring is anticipated to be a beneficial step in the right direction to significantly increase the frequency of monitoring, more readily understand developing conditions in the tailings dam structure, and in turn develop further knowledge on the phenomena and ways that the risk can be mitigated.Until a better understanding is gained and agreed in industry on how the liquefaction failure mechanism occurs and how/if this can be monitored, as opposed to only what might lead to liquefaction, it was not deemed appropriate to include this advice as a certainty in this research.Fell et al. (2015) describe that: there is a generally accepted principle that the level of monitoring and surveillance appropriate for a dam depends on the consequences of failure of the dam, whether the dam is being filled for the first time or is in general operation, and whether abnormal behaviour has been detected.
For each of these methods aside from direct observation, automated data collection is possible.When comparing to the opportunities presented by automated data collection, instrumentation that is monitored manually exhibit shortfalls including: . Increased time between readings, potentially missing the development of deterioration trends, or falling out of compliance based on a single, missed reading; .Increased labour demand to read and process the measurements; .Possibility for human error at the multiple touchpoints in the process, including reading the instrument, documenting the reading, processing the reading, or during other data handling activities; .Remote monitoring is limited to traditional survey, which either means that additional intrusive instrumentation needs to be installed, increased survey frequency is required, and the extent of surface monitored is limited to discrete points (with potential interpolation between these points); .Frequent installation of instrumentation to understand different areas can increase costs; .Less opportunity to identify when instrumentation is faulty or out of calibration.This may take several readings (at a reduced relative frequency) to identify the error; . Training of a higher number of field personnel to check, understand, and read instrumentation.Also trusting the professional interest of personnel undertaking the reading for reliable and accurate measurement; and .May be impractical for remote sites where personnel do not frequent.
Noting these shortfalls, there are a number of traditional instruments that are not suited to be upgraded to real-time monitoring.This is typically as a result of their installed configuration or the method by which the measurement is undertaken.Such examples include: . Water level gauge, limited by its nature as a static measurement attached to infrastructure; .Observation well, limited by its nature as an open hole unless combined with a piezometer; .Traditional survey, limited by the requirement to have an operator undertaking and manoeuvring the survey; .Catch containers, limited by its nature as a coarse water collection and measurement tool; .Magnetic extensometers, based on their measurement method relying on the sound of a buzzer on a tape reel prompting the taking of a measurement; and . Thermotic/ self-potential geophysical surveys, limited by the requirement to have an operator undertaking and manoeuvring the survey.
It is noted that in planning and establishing a monitoring system, the objective should remain as using the best tools for the intended purpose.If trained personnel readily frequent an area of interest, there is an inherent practicality on relying on visual observation, validated by instrumentation (potentially manually read, tailored to the risk of failure), as opposed to full reliance on a dense grid of automated instrumentation.Hence, it is suggested that a combination of data collection methods can be used, yet emphasised that the availability of newer (and increasingly more affordable) technologies can supersede many of the traditional instrumentation and monitoring techniques.

Real-time instrumentation
A detailed insight into the instruments that currently have real-time capability are summarised in Appendix A. The key performance metrics addressed by this instrumentation are extracted from Clarkson and Williams (2019) as These performance metrics are addressed due to their role in identifying and understanding the most prevalent modes of failure of tailings dams.These failure modes are described in the following sections, alongside a generalised rating of the performance metrics' usefulness in identifying the different modes.
It is important to note that the focus on automated instrumentation does not negate the importance of retaining the interpretation of collected data: the responsibility of which should remain with the engineer.The real-time automation of instrumentation is encouraged in the collection and processing of raw data to provide useful information, ready for interpretation.

Phreatic surface/water level
The pond water level of a tailings storage facility typically has a significant influence on the behaviour and trend of the phreatic surface.By measuring and understanding the structure's water balance, monitoring the pond water level as a recharge point, and the trend of the phreatic surface through the structure as a driven by the pond water, saturated tailings deposition, and environmental conditions, two loading scenarios can also be better understood: static/ normal operation (poor water management) and hydrologic (flooding, unpredicted reservoir water levels).

Applicability
A generalised rating of the usefulness of direct monitoring of the phreatic surface for identification of different modes of failure is presented in Table 1.
For direct and supporting indicators, justification is provided as

Direct indicators
. Overtopping: Excess rainfall and/ or uncontrolled reservoir water can cause overtopping; .Seepage: Critical reservoir level modelled for stability analysis defines allowable limit of phreatic surface and associated seepage flow; and .Slope instability: Increased pressure on embankment.At historic low reservoir water levels, upstream slope instability and internal deformation could be induced.

Supporting indicators
. Foundation failure: Saturation or lack thereof at the embankment toe can alter material stress states; .Internal erosion and piping: Strong evidence to show that failures or incidents occur at or above high historic reservoir levels (Fell et al. 2015); and .Seismicity: High degree of saturation can heighten susceptibility to liquefaction.

Instrumentation
In order to understand the behaviour of the phreatic surface/ water level, a number of instruments are available with real-time measurement capability.These are described on pages 1 and 2 of Appendix A.

Pore water pressure
Pore water pressure describes the pressure of water in voids between soil particles, or in discontinuities in rock.The phreatic surface generally defines zero (atmospheric) pore pressure, with the hydrostatic water pressure increasing linearly with depth below this.However, soil suction in fine grained soils can also cause capillary rise and negative pore pressures above the water table can be present.A change in pressure can cause an imbalance in the driving and supporting forces of a structure, resulting in destabilisation of varying scales (from particle transport to slope slumping).

Applicability
A generalised rating of the usefulness of direct monitoring of pore pressure for identification of different modes of failure is presented in Table 2.
For direct and supporting indicators, justification is provided as:

Direct indicators
. Foundation failure: 'Blow-out' or 'heave' occurs where a zero effective stress condition exists; .Seepage: Higher pressure could indicate a higher flow rate; and .Slope instability: Increased pore pressure on embankment.

Supporting indicators
. Internal erosion and piping: Pore water pressure change is typically a result of internal erosion and piping, meaning a measured change could indicate internal erosion has occurred; and .Seismicity: Fluid injection can increase pore pressure and cause induced seismicity.Pore pressure increases in active faults (seepage, monsoons, or induced) can also amplify effects.

Instrumentation
In order to understand the behaviour of the pore pressure, a number of instruments are available with real-time measurement capability.These are described on pages 2 and 3 of Appendix A.

Seepage flow
'Seepage data is one of the best indicators of dam performance' (Fell et al. 2015).Seepage processes describe the flow of water through the embankment, potentially giving rise to instability through piping (material  transport), slope instability and foundation heaving (increased pore pressure), or excess water losses (environmental damage).While the drivers are inherently linked to the behaviour of the phreatic surface and pore water pressures, additional approaches related to measurement of the flow of seepage can contribute to an understanding of the failure mechanism.

Applicability
A generalised rating of the usefulness of direct monitoring of seepage flow for identification of different modes of failure is presented in Table 3.
For direct and supporting indicators, justification is provided as:

Direct indicators
. Foundation failure: Irregular seepage can have an influence on the foundation saturation and strength (compressibility, stability) under and downstream of the dam; .Internal erosion and piping: Concentrated seepage in the foundation and embankment can influence vertical and horizontal flow gradients; and .Slope instability: Slope instability could be induced through excess material being displaced through seepage, undercutting or degrading the structural integrity of the slope.

Supporting indicators
. Seismicity: Increased seepage flow could heighten saturation of embankment, which in turn could heighten susceptibility to liquefaction

Instrumentation
In order to understand the behaviour of the seepage flow, a number of instruments are available with real-time measurement capability.These are described on page 4 of Appendix A.

Deformation and movement
Deformation and movement can be of varying scale and behaviour.Each observation deserves assessment and investigation into the root cause, considering that movement is not typically within the plan and design of the structure.Deformation can either be on the surface, or internal to the embankment, with behaviour typically any of vertical, horizontal, rotational, or translational.

Applicability
A generalised rating of the usefulness of direct monitoring of deformation and movement for identification of different modes of failure is presented in Table 4.
For direct and supporting indicators, justification is provided as:

Direct indicators
. Foundation failure: Movement indicative of foundation failure can be identified at and just beyond the toe of the embankment; and .Slope instability: Increased rates of deformation or settlement.
regional susceptibility to seismic behaviour is often understood and accounted for in the design.Regardless, monitoring techniques are employed to help understand the magnitude, distance to source, and the potential influence that these natural events may have on current activities.Mining activities have also been empirically proven to cause induced seismicity, from triggers such as underground rock burst, oil and gas extraction, fluid injection and hydraulic fracturing, and pore pressure increase in faults.

Applicability
A generalised rating of the usefulness of direct monitoring of seismicity for identification of different modes of failure is presented in Table 5.
For direct and supporting indicators, justification is provided as:

Direct indicators
. Internal erosion and piping: Resultant cracking within the embankment can initiate internal preferential erosion and piping; and .Slope instability: Can cause settlement or lateral spreading.

Supporting indicators
. Foundation failure: Depending on foundation materials, either static or dynamic liquefaction of saturated or partially saturated soils can occur.The stiffness and shear strength of a material is significantly reduced due to rapid increases in loading; and .Overtopping: Settlement as a result of particle rearrangement can reduce allowed freeboard.

Instrumentation
In order to understand the behaviour of the seismicity, a number of instruments are available with real-time measurement capability.These are described on page 7 of Appendix A.

Earth pressures
Earth pressures within tailings dams can provide an indication of the magnitude and direction of stresses, the percentage contribution of water and soil to total pressures (when combined with a piezometer), and in turn a comparison of design/ expected conditions against actual.If the stresses are different than anticipated, an anomaly may exist in the structure which may not have been accounted for in design, and hence in assessment and management of any associated risk.

Applicability
A generalised rating of the usefulness of direct monitoring of earth pressures for identification of different modes of failure is presented in Table 6.
For direct and supporting indicators, justification is provided as:

Direct indicators
. Foundation failure: Earth pressures can increase in the foundation indicative of uplift, heave or blowout, or increased/ excess loading from the embankment, tailings, and water regime combination; and .Slope instability: A change in lateral earth pressures, increasing through driving forces such as water pressure build-up or decreasing through relaxation or separation in the embankment body can indicate instability.

Instrumentation
In order to understand the behaviour of the earth pressure, a number of instruments are available with real-time measurement capability.These are described on page 7 of Appendix A.

Climate
Climate variability has primary influence on the water balance of the tailings storage facility.In the geotechnical sense, water can influence each of the different modes of failure to some degree.It is important to understand the amount and rate of change to compare against designed state and anticipate upcoming challenges.

Applicability
A generalised rating of the usefulness of direct monitoring of the climate for identification of different modes of failure is presented in Table 7.
For direct and supporting indicators, justification is provided as:

Direct indicators
. Overtopping: Heavy rainfall, snowmelt, and other extreme climatic conditions have been linked to historic tailings dam failures.Often, the introduction of additional, unpredicted inflows with a coincident high reservoir level can induce overtopping.

Supporting indicators
. For all: An increase in the anticipated water recharge (increased saturation) can raise the phreatic surface, induce additional pore pressures, alter material parameters and states, and in turn initiate any of the different modes of failure.
A decrease in the anticipated water recharge (drying) can induce cracking, shrink reactive soils, and alter the structure of the soil materials used to form the embankment.

Instrumentation
In order to understand the influence of the climate, a number of instruments are available with real-time measurement capability.These are described on Page 8 of Appendix A.
There are two main challenges to highlight when considering online monitoring systems.The first is internal and external data security.Data security is being addressed through software inclusions such as fine grained user permissions, allowing in-house managers to designate user control and viewing levels, and secure communication protocols.The second challenge to highlight is ensuring that processed data should not be mistaken for interpreted data.'The delicate task of fine interpretation belongs to the engineer' (Fell et al. 2015).The measurements need to be applied to the tolerable levels at the instrument location to translate the data into an understanding of dam performance.

Conclusion
This research established a catalogue of real-time instrumentation and monitoring techniques for tailings dams, focusing on the key performance parameters for identification of prevalent failure modes and mechanisms.The advancement of technology to be able to present measured data in realtime mitigates many of the risks associated with manual data collection.Through mitigation of these risks, the responsible engineer can more readily and reliably understand the performance of the tailings dam.
Through collaboration with a global selection of instrumentation suppliers and a generalised approach (at this stage of the wider project), it is anticipated that this catalogue is applicable to tailings dams across the globe.By transparently listing the advantages, disadvantages, specifications, and considerations that are pertinent to each tailored instrument type, practitioners have the opportunity to be better informed when engaging with consultants and suppliers, but most importantly are presented with a tool that can help toward a better understanding and management of their tailings dam.

Table 1 .
Rating of usefulness of phreatic surface monitoring for identification of different modes of failure.

Table 2 .
Rating of usefulness of pore pressure monitoring for identification of different modes of failure.
Note: Direct = Direct Indicator, Supporting = Supporting Indicator, Other = Recommend using different parameters.

Table 5 .
Rating of usefulness of seismicity monitoring for identification of different modes of failure.

Table 6 .
Rating of usefulness of earth pressure monitoring for identification of different modes of failure.Direct Indicator, Supporting = Supporting Indicator, Other = Recommend using different parameters.