Integration of satellite and terrestrial monitoring methods to identify the connection between triggering factors temporal variation and changes in large slope instabilities activity rates

Abstract

Recent years have seen significant advances in landslide monitoring technologies. These systems can be broadly subdivided into “in situ” and “remote sensing” types. Despite the continuous improvement of both types, the first are representative of the “traditional approach” while the second entails more recent technologies (e.g. satellites and drones). Numerous researchers have modelled displacements of large active landslides in response to triggering factors, mainly using traditional techniques monitoring. At the same time, space-borne data have been increasingly used in such studies, and in particular, Interferometric Synthetic Aperture Radar (InSAR). Nevertheless, InSAR methods show advantages and limitations, and traditional in situ techniques are still in everyday use by geotechnical engineers. Moreover, single landslides frequently require specific combinations of the aforementioned approaches in order to suit the unique geologic, geomorphic, and hydrologic character of the given area and, thus, the interpretation of slope monitoring data for establishing an early warning system remains largely subjective. The general objective of my PhD project is to contribute to the field of monitoring large slope instabilities by exploring the potential to use InSAR techniques to detect changes in slow-moving landslides caused by temporal variations in their triggering factors. The motivation for my work is that InSAR techniques offer highly favourable cost-benefit ratios in landslide monitoring and the constant improvements in the technology offers new opportunities for managing landslide risk. I selected three study cases in Northern and Central Italy, with different geological and geomorphological contexts: 1) the rockslide of San Vito Romano, situated in a landslides-prone basin in an area of high rainfall; 2) the Deep Seated Gravitational Slope Deformation at Fiastra, affecting limestone bedrock in a seismically active area; and 3) the Deep Seated Gravitational Slope Deformation at Maso Corto, located in an alpine environment. I evaluate the variation through time of landslides activity, as detected by both traditional and remote sensing techniques, and I relate it to the changes in the intensity of possible triggering factors. Moreover, I performed statistical analyses on the satellite SAR time series and compared the results with those obtained using ground monitoring systems. In addition, my research also attempts to identify more poorly understood triggering factors such as permafrost degradation. My final objectives are to provide future perspectives on these techniques, to summarize their limitations, and to provide insights for planning early warning systems. In San Vito Romano, we performed a geomorphological field survey to define the landslide area. We then incorporated geotechnical data of 25 boreholes, 13 inclinometers, 20 piezometric, a rainfall dataset, and 6 extensometers. Moreover, the average monthly and annual velocities along the sliding surfaces intercepted by inclinometers have been calculated. We then analysed PSInSAR data provided by the Envisat satellite and the Cosmo Skymed satellite performing statistical analysis on the time series (calculation of characterising periods and cross-correlation with rain for a period of less than one year). Finally, we calculated rainfall thresholds for shallow landslides in the study area. The study shows that the study area is being affected by a translational rockslide that transforms into a rotational slide and which deposit increases downslope up to 30 m deep. The annual average velocity along the sliding surface never exceeds 5 mm/y. This value is consistent with Envisat and Cosmo SkyMed surface data. Within the area of the landslide, the groundwater table is shallow through most of the year. However, it is influenced by rainfall and, at the beginning of summer, it falls rapidly. We have identified appreciable deceleration during the summer months along the most active sliding surfaces. Thus, we consider that the reactivation of the landslide occurs during the autumn-to-spring period and that it is driven by the progressive accumulation of water from rainfall. Furthermore, when there is no movement, we infer that the corresponding water level depths represent threshold values below which the landslide is relatively stable. SAR data might provide information on water level changes when piezometric data are not available but did not allow us to further characterize landslide activity in our study case. Finally, we identified 200 mm over 3 days and 250mm over 10 days as rainfall thresholds for shallow landslides by performing a statistical analysis. We consider that this work provided several insights that can be used by public officials to reduce landslide risk in San Vito Romano village. In Fiastra we documented the pre-seismic evolution of a Deep Seated Gravitational Slope Deformation and measured its seismically induced displacements during the earthquake sequence of Central Italy in 2016 and 2017. A multidisciplinary approach that combines a field geomorphological survey, installation of permanent GPS stations, and DInSAR measurements was adopted for this study. The study shows that the seismic reactivation of the Fiastra Deep Seated Gravitational Slope Deformation mainly depended on the magnitude of the earthquake and the distance from the epicenter, and only secondarily on the number of big earthquakes on a given day. We measured about 60 mm as the maximum instantaneous displacement, which occurred during an Mw 6.5 earthquake around 25 km far from the study area. Earthquakes smaller than Mw5 or farther than 70 km from the study area did not produce significant movements. For earthquakes larger than Mw5, we found a linear correlation between the measured displacement and the combination of magnitude, epicentral distance, and the number of earthquakes per day, assuming the same hypocenter depth (7-10 km) for all events. The step-like reactivations triggered by earthquakes, are one to ten times higher than the more linear, <5 mm/y average normal displacement rate of the deformation, which is associated with creep-type motion and caused by river erosion of the toe of the slope, although the influence of rainfall on movement rates cannot be completely ruled out. The Maso Corto case presents the results of an experimental composite monitoring of geomorphic processes carried out within the project “SloMove”. The study aimed: 1) to understand the state of activity of geomorphological processes in the study area, in particular slope instabilities and rock glaciers, in order to 2) reconstruct the probable state of permafrost in the area and hence 3) to provide insights into the role of permafrost as factor influencing large slope instabilities. To successfully achieve these goals, we determined activity rates of the aforementioned geomorphic processes using a multidisciplinary investigation approach that integrates within a GIS a structural-geomorphological survey, GPS measurements, and time series analysis of PSInSAR data. By coupling all these techniques, it resulted that the rock glaciers in the study area are active with very low movement rates compared to others in the Alps. Based on this, we assumed that discontinuous, thin, and shallow permafrost may be present in the study area and that, although it is probably degrading, the seasonal variation of its active layer can still trigger the rock glaciers movement. Moreover, the study shows that a ~2km2 area north of Maso Corto is affected by a Deep Seated Gravitational Slope Deformation moving at considerably faster (1-3 cm/y) than most other similar phenomena in the Alps. Both SAR and GPS data, indicate that the deformation is active throughout the year, suggesting the presence of a constantly acting causative factor. Debuttressing cannot alone account for such high activity. We also have to exclude seismic activity as a trigger, because, in the entire monitoring period, no M>4 earthquake happened close to the study area. At the same time, we think that, especially in summer, rain and snow melt can induce seasonal acceleration phases. Nevertheless, considering our inference about the state of permafrost in the study area, the progressive degradation of permafrost might explain the movement rates better than other aforementioned possible triggers. This thesis offers conclusions on the geomorphological aspects related to the phenomena of large landslides and Deep Seated Gravitational Slope Deformation, and the influence of the identified triggering factors on their activity rate. The study of the San Vito Romano study case, providing rainfall thresholds for the triggering of shallow landslides and indicating ground water levels above which deep the rockslide that affects the village can be reactivated, furnished insights can be used by the local stakeholders to manage landslide risk. The Fiastra and Maso Corto case studies tell us that other Deep Seated Gravitational Slope Deformation in the European mountains may currently be active with high activity rates and provide insights on the response of these phenomena to large earthquakes as well as the to the contemporary permafrost degradation. These findings can be compared to similar study cases of large slope instabilities. Moreover, my research also indicate the main advantages and drawbacks of the method used to accomplish the objectives of this work. The average velocity of a landslide is the first factor to consider when plan an appropriate monitoring system, as demonstrated in the San Vito Romano case study. Here, the movement seasonal variations are too small (around 1 mm or less) and, for this reason, it was not possible to relate them to the significant changes through time of the intensity of its trigger (i.e. rainfall). In contrast, in the case of the Fiastra and Maso Corto examples, the periodical acceleration and deceleration phases were larger and easier to intercept. In addition, the Fiastra study demonstrated that as the mean velocity of a landslide decreases, more frequent measurements of its displacement are required. Here, due to the high frequency of GPS measures, it was possible to document seasonal variations in movement of <1 mm. Related to this point, the length of the time series is also an important parameter to be considered when investigating the mid-to-long term evolution of a landslide. The Radarsat dataset proved to be the most useful for balancing the frequency of measurements (24 days) and the length of the time series (3 years) in relation to the rate of the monitored processes. Another issue in the use of the aforementioned techniques was the necessity to relate the measured movements at the surface to deeper movements of the slide mass. In this regard, the underground data at San Vito Romano proved to be pivotal in understanding the landslide. However, in the Fiastra and Maso Corto cases, which lacked subsurface data, the signal stemming from shallow processes demanded detailed interpretation to isolate the deep. Another important consideration is that triggering factors should be measured in situ, since systematic errors can be introduced by collecting data on triggering factors far from the site of interest. The last observations pertain to the statistical approaches used in this work on time series (cross-correlation, periodicity calculations, down sampling, seasonal mean velocity calculations, coupling time series from different points) which demonstrated useful in order to provide further details on acceleration and deceleration phases of the landslides in response to variations in the intensity of the triggering factors. In conclusion, with the study cases presented in this thesis, I hope to have contributed to understanding the triggering factors for large slope instabilities and to have provided further insights into the applicability of monitoring methods, underlining that the integration of remote sensing, in situ techniques, and field surveys can still be seen as the best path forward.