The development and the use of the OM started in the 1940s. The historical evolution of the use of the OM started with Peck (1969) and Terzaghi et al. (1996). The OM had the aim to reduce the costs during construction that were caused by designing earth structures based on the most-unfavorable assumptions. Instead, the design according to the OM is based on the most-probable conditions rather than the most-unfavorable. Gaps in the available information are filled by observations: in this case by the monitoring data. These observations aid in assessing the landslide evolution, checking its velocity trend with respect to the ongoing works, and the behavior of the structure during construction, which can then be modified in accordance with the findings.

Usually, monitoring has the aim of verifying the functionality of the works over time, observing the relation between the design plan and the phenomena evolution. When coupled with the OM, the goal of monitoring, which plays an active role in both design and construction, is also to validate the adopted design solution or to identify the most appropriate design solution among the planned ones. The OM facilitates design changes during stabilization works and establishes a framework for risk management.

Peck discussed the OM in his 1969 Rankine lecture (Peck, 1969) and defined two OM approaches: a.

ab initio approach, adopted from inception of the project;

GB-InSAR displacement measurements were checked with a frequency proportional to the velocity of the landslide. During the emergency phase, specialized personnel was permanently at the monitoring center to provide continuous data check and validation. In the later months, during the most stable periods, data were still checked at least daily. Empirical velocity thresholds were implemented to define different warning levels. The results of the monitoring were dispatched to the stakeholders in the form of daily monitoring bulletins. Extraordinary bulletins were issued in case of sudden accelerations. Bulletins were used as support for decision making, within the framework of the OM.

When monitoring data review showed that the predefined design exceeds predicted condition or a velocity threshold is exceeded, this would then trigger a “recovery stage”. This consists in implementing emergency planned measures to secure the safety of the site staff and the general public while the unexpected event is fully investigated. Any decision in this stage need to be made rapidly; for this reason the assessment will be qualitative rather than quantitative. Once the safety of the site has been secured, the project team has to bring the project back to a fully stable condition, which means first carrying out a “design review stage” comparing the actual conditions with the original design.

The GB-InSAR technique employs a radar system, which is composed of a coherent microwave trans-receiver unit based on SAR (Curlander and McDonough, 1991) and on interferometric techniques (InSAR), originally developed for earth observation from satellites (Zebker and Goldstein, 1986). The synthetic aperture is formed by moving the radar head along a linear rail.

The GB-InSAR system radiates microwaves on the investigated area to measure the backscattered signal, obtaining an SAR image and providing measurement of ground displacement through remote sensing. The system detects a continuous 2-D areal distribution of the occurred deformation within the observed ground portion, without the necessity of positioning targets on the ground and without any physical contact with the slope (Tarchi et al., 1997). The obtained SAR image is created based on the spatial resolution along the direction perpendicular (range resolution, ΔRr) and parallel to the rail synthetic aperture (azimuth, or cross-range resolution; ΔRaz) (Luzi, 2010); each pixel in the image contains amplitude and phase information of the investigated scenario's backscattered echo.

By comparing the phase information between two radar acquisitions of the same scenario, it is possible to obtain a 2-D displacement map of the investigated area that constitutes an interferogram (Luzi et al., 2004; Monserrat et al., 2014). One of the limitations of this technique is that only the displacement component parallel to the line of sight (LOS) can be measured; thus, displacements that occur along a direction perpendicular to the sensor are missed, making determination of the location of the installation point crucial (Fig. 5a).

Regarding the investigated site, the radar system was installed to ensure the sensor LOS is as parallel as possible to the expected direction of the landslide motion. The high precision (sub-millimeter) of the technique contributes to minimize this limit, detecting even small displacements. Other important limitations of the GB-InSAR technique are due to temporal decorrelation and atmospheric noises (Luzi et al., 2010).

On 29 April 2010, the apparatus was installed on the opposite slope located at approximately 4 km from the earthflow.

To optimize the monitoring system efficiency, the radar should be placed at a stable location, with the minimum presence of obstacles between the radar sensor and the investigated objects. The technical features of the GB-InSAR campaign at Montaguto are shown in Table 1 and in Fig. 5b. To detect both short- and long-term movements, interferograms with a temporal baseline spanning from 3.5 min to 1 month have been used.

During the GB-InSAR monitoring, data were transferred via ftp to the processing and post-processing unit for the generation and production of the interferograms (Fig. 6) and the monthly cumulated displacement maps (Fig. 7). Negative displacement values indicate a movement toward the sensor (shortening along the LOS), whereas positive displacement values indicate movement away from the sensor (lengthening along the LOS).

Based on the results of the landslide velocity pattern, the DPC, with the operational support of the Italian Army, started stabilization efforts using large earth-moving machinery along the toe of the earthflow. As a result of the monitoring and subsequent stabilization efforts, the railway was reopened on 7 June 2010 and the state road 90 was reopened on 10 July 2010.

The GB-InSAR system was one of the key elements in the work planning; worker safety was improved because alerts could be provided regarding sudden accelerations that required stoppage of work or evacuation.

During the first stages of the monitoring activity, for few weeks, the GB-InSAR instrument was partnered with a webcam and with an infrared thermal camera (IRT). A comparison between the optical images and the interferograms was crucial for the interpretation of the radar images and therefore the detection of unstable areas, especially of the landslide toe portion (Fig. 6), while the IRT allowed a very accurate investigation of the water flow paths, wet areas and drainage directions (Fig. 8). In an OM perspective, the individuation of wet and fast-moving areas (roughly coinciding) directed the design of the remediation works and allowed us to give priority to such particular areas and to all the interventions aimed at draining water from the landslide. This first emergency phase ended in July, when the displacement rate decreased as a result of the constructions and the dry season (Fig. 10).

The toe of the landslide is the sector that experienced the highest detected velocity during the emergency phase (Fig. 6). The average displacement rate reached approximately 1.0 m day-1, although there were significant intra-day fluctuations. The maximum recorded velocity was measured at 2.9 m day-1 on 1 June 2010.

To understand the kinematics of the entire landslide toe, four sectors were identified during the emergency phase (A, B, C and D) (Fig. 6) to represent the main reference points for assessing landslide evolution. Aiming for an operational purpose, such sectors have not been identified based on geomorphological criteria but only on their displacement velocity, as visible on the interferograms (Fig. 6). To better follow the temporal evolution of each sector, a representative pixel has been selected within each of them; hence displacement time series relative to these pixels have been obtained.

During the months following the emergency phase, two more unstable areas were identified, labeled sectors E and F (Fig. 7). From the beginning, the data processing has been subjected to many variations induced by the landslide evolution.

In the initial monitoring days, the velocity of the landslide was so high that the interferograms needed to be calculated over a very short time interval (around 4 min) (Fig. 6).

When the velocity started to decrease, the displacement cumulated in 4 min was lower than the sub-millimeter sensitivity of the GB-InSAR and therefore displacements were only detectable on 4 h interferograms (Fig. 9.1–2) and, later, on 24 h interferograms (Fig. 9.6). In March 2011, the E sector is very clearly detectable even in the 4 h temporal baseline (Fig. 9.8), and interferograms in Fig. 9.8–9 show the progress of the works implemented for the Rio Nocelle water channeling and the upper-medium sector works. Finally, starting from March 2011, with the aim of making the monitoring activities even more suitable for landslide evolution analysis, monthly cumulated maps were processed for the entire acquisition period (a monthly cumulated maps time laps is available in the Supplement). The elaborated cumulated displacement maps are represented with a color scale for visualizing (i) stable areas in light green, (ii) areas characterized by displacement towards the sensor LOS with colors from yellow to red and purple (maximum cumulated displacements) and (iii) areas characterized by displacements away from the sensor LOS with colors from dark green to deep blue.

Some monitored sectors have slowly continued to deform. For example, the D sector and a strip upstream in sector A continued a slow movement from December 2010, whereas the E and F sectors started their slow and intermittent movement in a later time. Movement in the E sector (located in the middle-upper part of the landslide) had movement detected in November 2010; in the F sector (the upper main scarp of the landslide) deformation was detected in December 2013.

From the data analysis and interpretation, it was possible to recognize and define the evolution of the observed phenomenon, in terms of a decrease and an increase of the deformation rate.

Because the stabilization projects were continuously checked and monitored by the GB-InSAR system and the draining and channeling works were efficiently carried out, it was possible to determine that the deformation processes and movement speeds were gradually decreasing (Fig. 10).

After performing an analysis of the results of recent surveys and direct observation of the area, as well as the results from the monitoring, a design for stabilization works was established that can be updated and adapted, depending on the landslide response to the stabilization efforts. In September 2010, the main work phases were outlined. This represents the first application of the OM, referring to the ab initio approach, previously mentioned.

The observations of the surface features produced by differential movement of the landslide and the analysis of the monitoring data enabled to understand the behavior of different parts of the landslide and to estimate a zonation of landslide elements; therefore, to better plan and conduct the required interventions, the earthflow has been divided into three zones. For each of them the most appropriate works were defined accordingly (Fig. 11).

Upper sector (depletion zone): lake drainage, upstream surface drainage (diversion ditches), re-profiling of the main scarp, modification of the slope profile and restoration of stream channels to their natural conditions.

The upper part of the landslide was characterized by the presence of a system of lakes, whose water was directed by drainage trenches and shallow channels into a well and then channeled into a watershed located outside the landslide (Fig. 11).

Shallow channels with hydraulic jumps were constructed in the middle part of the landslide. Deep drainage trenches that allowed deep groundwater to spring outward at the sites of hydraulic jumps were an integral part of the design.

The construction of shallow channels coupled with deeper groundwater drainage trenches was repeated at the lower part of the landslide. The water from the lower part and from lateral channel system is conveyed towards a natural watercourse that flows beyond the toe of the slide (Fig. 11).

The GB-InSAR monitoring data analysis was essential for the location and effectiveness assessment of the lower one, whose stabilizing effects were clearly visible to the radar apparatus (Fig. 12). In this manner, the abundant stagnant water within the surface depressions that had been created by the landslide was eliminated, thus reducing water infiltration and the rate of movement.

The stabilization works involving gabions and profiling of the landslide started at the toe (Fig. 13, blue box), where the interventions for water drainage were completed, with the goals of protecting the state road and the railway and restoring their use.

First, steel-reinforced gabions were installed to build retaining and permeable systems of considerable size (lower sector). Then, the landslide was reshaped in accordance with the constructed drainage works (middle sector). The location and timing of the interventions were selected based on the monitoring data results (Fig. 14).

Regarding the middle sector, a very interesting example of the coupled integration between GB-InSAR and work design and execution is the analysis of the landslide behavior in the mid-sector, i.e., the so-called “elbow” (E sector): here, the movement increased during November 2011–May 2012 despite the execution of the drainage works (Fig. 14). The activity and evolution of this sector was promptly highlighted by GB-InSAR data; the monthly displacement cumulated maps refer to the two highest deformation rate periods (March–April 2012), when the velocity reached approximately 1 cm day-1. To verify this information, some field surveys were performed that showed the damage to three weirs and the geotextile filter (Fig. 13b). Based on these observations, a design variation involving additional drainage elements in the main channels (Fig. 11, no. 10) was incorporated into the design (Fig. 12). This variation constitutes a valid example of the “best way out” approach, as suggested by Peck (1969).

To highlight the project's efficiency in relation to the slowing rate of displacement, a velocity graph was constructed to compare the beginning and end of the different stabilization works affecting the most unstable area, identified as the E sector (Fig. 15). In May 2012, most of the stabilization works were completed.

Because of the high acquisition rate and accuracy of GB-InSAR data, the efficiency of the undertaken activities was continuously verified through the OM by observing the time series of the velocity recorded at critical points (Fig. 16i).

The prosecution of monitoring verified the long-term effectiveness of the stabilization works, which prevented reactivations even after rainy periods (Fig. 16iii) (with a small acceleration of sector D during early autumn 2010 being the only exception).

The GB-InSAR was also used as an early warning tool. During the very initial phase of the emergency, when the earthflow displacements were high, a warning velocity threshold was set at a level of 2 m day-1. The threshold was exceeded on the 10, 14, 24, 25 and 31 days of May 2010; during these days, people involved in interventions were recommended to take particular care during the execution of the works located at the landslide toe area (sectors B, C and D) (Fig. 16ii).

The Montaguto landslide thus represents a very interesting example of hydraulic and structural engineering techniques in stabilization work focused on minimizing the environmental impact, the execution and effectiveness of which have been driven and evaluated by accurate monitoring data.