1. Introduction
Land displacements such as mudflows, landslides, topples, or slope failures are triggered by the destabilization of a slope through rainfalls, seismic events, changes in water levels, or human activity, among others. For instance, mining activities such as excavation, blasting, material removal, and water extraction can trigger such events that may compromise worker safety, mine stability, equipment, surrounding communications, and power infrastructure. Landslides result in extensive damages to the environment and infrastructures, and in thousands of lives lost every year, according to the US geological survey [
1]. The impact of mining on land stability is significant [
2]. Nevertheless, mining is an important component of the global economy. Forty companies share the vast majority of the global mining revenue. In 2019, this represented USD 692 billion [
3]. Reducing mining activities is not an option, so reducing their impact of landslides by enabling effective, secure, and qualitative monitoring is essential.
Several methods have been used for slope monitoring and landslide measurements. Terrestrial laser scanning (TLS) and global navigation satellite systems (GNSS) (including global positioning systems (GPS)) are two common techniques used for this purpose [
4]. Another one is the synthetic aperture radar interferometry (InSAR) [
5,
6]. However, land displacements are mostly monitored using a series of sensors disseminated on the ground surface [
7]. To detect land motion, or measure its rate, magnitude, and direction, monitoring is conducted using devices such as slope movement sensors (extensometers), accelerometers, inclinometers, tiltmeters, distometers, prisms, survey stations, and vibration sensors [
8]. Internet of Things (IoT) monitoring systems are often used for geotechnical information, alone or in combination with innovative technological elements [
9,
10,
11]. Traditional manual reading is being progressively replaced by the deployment of wireless networks of commercially available instruments to collect, transmit, and process land displacement data, as displayed in
Figure 1 for a mining environment. Such monitoring networks have several advantages:
They do not require human interaction to collect the data;
They require minimum maintenance;
They are battery powered;
They have low power consumption;
Connectivity to the internet facilitates real-time data visualization and further analyses.
Despite these advantages, however, motion sensors provide only discrete information limited to the footprint of the sensor, and a specific time range, resulting in potential information gaps. A considerable amount of work in the field of monitoring has lately been focused on the modeling and forecasting of displacements, to improve monitoring systems while reducing their cost. These predictive systems often exploit the initial geotechnical model of the structure and the data collected by IoT devices is used to update the model and calculate the safety factor of the structure. In particular, PLAXIS 2D and 3D are powerful and user-friendly finite element packages intended for two-dimensional and three-dimensional analyses of displacements and stability in geotechnical engineering and rock mechanics. Nevertheless, it remains generally difficult to detect the long-time precursors of such events. The InSAR technique can be used to identify such precursors, since it allows for analysis over a wide area and a long time period of the trend of slope displacement, velocity, and acceleration, which are the best indicators of ongoing failure processes [
12]. Photogrammetry processed aerial photography and the Light Detection and Ranging (LiDAR) survey are also frequently used. Here, we report on a flexible commercial framework, which can orchestrate diverse hardware products and software modules, addressing both large scale and in situ aspects of monitoring.
4. Discussion
The data collected from each of the sources were compared with the visual inspection of both the events observed in June and July 2020. A strong correlation was reported between the actual events and the robotic total stations information since the prisms were installed on the hill slopes themselves. False positives and false negatives were not observed, while the direction of the vector of the displacements matches the expected direction, given the contour lines.
2020 was the warmest year on record in Lithuania. The average annual temperature measured was of 9.2
C. This was 2.3
above the multi-annual average. The year 2020 was also relatively dry in Lithuania since the total amount of precipitation was 7% less than a normal year [
34]. June and July are typically two months with heavy rainfall in Vilnius, which are condensed in few days. In particular, in 2020, there were several days of heavy rain in the beginning of June, just before the slope failure event observed. During July, the weather was mainly dry, although heavy rains were reported on the same day of the slope failure event. These rains could have triggered the specific events, although the general trend displayed in
Figure 11 and
Figure 12 demonstrates that such events are to be expected.
Since the tiltmeters are located on the castle walls, they measure the effects of the incidents on the structure, rather than the incidents themselves. The Loadsensing data show minor changes during the June 2020 event in the tiltmeters located in the same direction away from the castle, while tiltmeters outside the region of interest (ROI) do not detect the event, as reported in
Table 4. The measurements were also reported in
Figure 17 and are consistent with the expected behavior.
The magnitude of the event, as reported by the tiltmeters in the ROI, has a smaller scale than the one from the total stations. This is also consistent with expected results since the tiltmeters are installed on a more stable structure (the castle itself, located at the top of the hill, and the castle walls located at the bottom of the hill) whilst the landslide event occurred on the slopes. The tiltmeters did not detect a trend in the movement but a sudden change, which is also consistent with the type of slope failure observed: a sudden landslide. This makes it difficult to detect trends or patterns that can be observed in other scenarios.
The tiltmeters did not detect the July event, as reported in
Table 5, even in the ROI, which demonstrates that, although the hill slopes have experienced a slope failure event, the castle itself was not affected. This illustrates the fact that the displacement of the surrounding area does not always pose a threat to the structure and infrastructure being monitored. Nevertheless, the long-term shift detected by InSAR is also a precursor of future slope failure events.
The results described in
Section 3.3 point towards the wide applicability of InSAR for the estimation of displacements in cases of slope instability and failure. The success of the approach appears to vary between failure events and specific scatterers. In this observation of medium resolution Sentinel-1 data, the coverage of the slopes by the InSAR measurements is not totally spatially consistent. This may be explained by one of the key assumptions of robust InSAR data being consistent and persistent reflection results over time. Furthermore, sudden displacements may lead to errors in the displacement estimation. As indicated by
Figure 16, the slope failure on the southwestern front was a large and sudden movement with a magnitude of displacement that is relatively difficult for this InSAR analysis (temporal and spatial resolution constraints) to capture. At the castle hill area of interest, the incidence angle and geometry of the slope compounded the challenges in correlating coherent signals with the event.
In comparison, the combination of InSAR with other technologies has been extensively validated, for instance with LiDAR [
35], GNSS [
36,
37], and other remote sensing techniques such as ground-based radar [
38]. InSAR has also been used to study slope failure events in open pit mines and show how the technique could be used to detect precursor signs of catastrophic slope failure [
36]. The combination of InSAR with 2D finite element models [
39] and 3D slope stability software [
40] has also been reported. Nevertheless, there are still few studies of the combination of InSAR with underground and terrestrial structural monitoring. Selvakumaran et al. have reported comparable InSAR and automated total station (ATS) readings [
41]. In this specific work, the comparison of InSAR and ATS readings turned out to be comparable regarding the relative movement of points along the bridge from one another, although there was no data correlation of InSAR with other types of in situ sensors. Lastra et al. reported a high correlation between GNSS and InSAR measurements without correlation with the extensometer measurements [
42]. The Earth Dam of Conza della Campania was monitored with a combination of extensometers and InSAR, showing a strong agreement between the displacements recorded by both monitoring techniques [
43]. Finally, the comparison of InSAR results with in situ monitoring by inclinometers was reported, which validates InSAR as a valuable technique to monitor landslide displacements [
44,
45].
The results proposed here demonstrate the complementarity of both techniques. This might be relevant for applications in mining monitoring, where slope failures frequently occur, especially for open pit mines, quarries, and Tailings dams. Future work will include the validation of the technology in real dormant and active mining sites, as well as the inclusion of other types of sensors relevant for mining monitoring, such as in-place inclinometers. Further steps for this work will also address the improvement of the metadata for the RTS and Loadsensing, mainly regarding location and orientation, and will be solved by working on different use cases. Agreements have been reached for new testbeds in an active open pit mine and a tailing storage facility. These new testbeds will also allow the validation of this method in real mining scenarios. Other steps might involve the development of numerical simulations of landslide failures to interpret the measured result [
46,
47], and the application of relevant aspects of the theory of mining area displacement.
5. Conclusions
The authors report here the successful implementation and validation of a commercial monitoring system which correlates the data from IoT data acquisition and monitoring system with other data sources (robotic total stations and the SkyGeo InSAR) in a real testbed. The data are correlated in one slope failure event, where all three technologies have detected displacements in the same directions, with different orders of magnitude consistent with the location of the sensors. In another event, the RTS and InSAR detected the event in a correlated manner whilst the tiltmeters demonstrated that the structure was not affected.
Furthermore, the complementarity of both technologies is demonstrated, since the InSAR data observe ground displacement on a large scale and over a large period of time, detecting displacement precursors before the slope failure events occurred, whilst the IoT system detects the actual consequences on the structure being monitored. The in situ sensors data are also used to optimize the InSAR data analysis.