Applications: Subsidence and Landslides (4)

An integrative assessment of hazards related to glacier lakes requires to focus on the glacier lake, its dam properties, and processes in the lake surrounding, including the position and fluctuations of the glacier tongue, as well as inclinations, potential displacements and thermal conditions of adjacent slopes. A transient perspective on these factors allows anticipating potential future developments. Despite these widely acknowledged dependencies in time and space, the integrated detection, monitoring and modelling of glacier lakes, surrounding slope instabilities, and glacier conditions and behaviour that potentially affect the glacier lakes has to our best knowledge not been done so far and represents a novel, though highly needed approach. In the framework of the S:GLA:MO (Slope stability and Glacier LAke MOnitoring) project, supported by the European Space Agency, we aim to provide to the organizations involved as users a first-order assessment of hazards related to glacier lakes based on EO derived products, in-situ data and modelling results with recommendations regarding potential future developments and follow-up monitoring activities. EO data (optical and SAR) are considered for the monitoring of glaciers (outlines and velocity fields) and glacier lakes (area), for the compilation of an inventory of landslides with indication of the state of activity, and for the production of Digital Elevation Models (DEM). In the first phase of the project, we focused on a demonstration of the service concept at two sites in Switzerland where a significant number of data was already available to us from other projects. In the second phase of the project the S:GLA:MO methodology is applied to glacier lakes in the Cordillera Blanca in the Peruvian Andes, the Gorno-Badakhshan Autonomous Oblast in Tajikistan, and South Greenland.

In this contribution we will discuss results obtained with regard to slope instabilities around glacier lakes in our sites in Peru, Tajikistan and Greenland. Satellite SAR interferometry, recognised to be a mature and well consolidated technology for the mapping and monitoring of land deformation, is considered together with geomorphological information from photo-interpretation for the compilation of inventories of landslides with indication of the state of activity. While the analysis of aerial photographs allows us to recognize many phenomena of instability and to define their limits, radar interferometry allow us to characterize the class of intensity of the landslides with a deformation time-series extended over 20 years. Various interferometric approaches, ranging from single wrapped interferograms for the characterization of the most rapidly moving periglacial landforms (rates of movements >1 cm/day) to short-baseline multi-temporal interferometry and persistent scatterer interferometry are considered. Analysis is performed using stacks of ERS-1/2, ENVISAT, TerraSAR-X, Cosmo-SkyMed and Radarsat-2 SAR data acquired until 2014. The topographic reference is taken from TanDEM-X pairs interferometrically processed to DEMs with support of the SRTM DEM. Although processing algorithms are continuously improving, we found that the quality of our results mainly depends on other factors, like availability of satellite SAR data, temporal interval of the SAR acquisitions, carrier frequency and spatial resolution of the SAR image, land cover (particularly presence of vegetation and snow cover), topography and availability of a good DEM, and spatial dimension and rate of movement of the phenomenon under analysis.

Space-borne Differential SAR Interferometry (DInSAR) is a well-established technique to monitor surface deformation caused by anthropogenic and natural processes. Nowadays, a large amount of SAR data acquired from different sensors are available, and forthcoming missions will further increase the opportunity of measuring ground displacements at high spatial and temporal resolutions. One of the known limitations of DInSAR approach is that measurements are taken along the sensor’s line-of-sight (LOS), thus in many cases only a percentage of the real deformation magnitude can be retrieved. By combining ascending and descending acquisitions of the satellite covering the same area, it is possible to decouple the horizontal (mainly in East-West) and the vertical components of the surface deformation, but 3-dimensional displacements are not accurately detectable when considering DInSAR measurements only (despite using three different tracks). However, the knowledge of the 3-dimensional deformation field is often a real need for a detailed interpretation of natural processes, as for example those associated to landslide phenomena.

In this work, we present the results obtained by exploiting DInSAR data in order to constrain numerical models of active landslide processes. We consider a DInSAR dataset relevant to the Tena Valley, Central Spanish Pyrenees, and we focus on the El Portalet ancient landslide, which is one of the test areas considered in the framework of the FP7 LAMPRE Project (http://www.lampre-project.eu/). First, ERS and ENVISAT data (ascending orbit) acquired in the 1993-2008 time period have been jointly processed to retrieve long-term surface displacement time series. SAR data have been processed using an innovative advanced PSI approach for maximizing the final density of measurements which is a key factor for improved interpretation and analysis in landslide applications (Albiol et al. 2014). Then, by considering the DInSAR observations as input, we optimized the parameters of a 2-dimensional Finite Element Model (FEM) for the analysis of the kinematical behavior of the landslide, which has been constrained through field data and geomorphological interpretations. The optimization was performed by following the approach proposed by Manconi et al., 2009, i.e. combining the ©COMSOL Multiphysics in a Genetic Algorithm scheme in order to retrieve best-fit model parameters that minimize the root-mean-square of the difference between measured and calculated displacements. Moreover, we have considered different constitutive models of the soil (e.g., elasto-plastic, poro-elastic, creep), in order to evaluate the most appropriate soil model for characterizing the landslide behavior. Secondly, the results of the optimization have been used to build up a 3-dimensional model, which allowed us to retrieve the 3-dimensional components of the surface deformation field (Tizzani et al., 2010). Thus, starting from a DInSAR dataset measuring along LOS only, we have retrieved 3-dimensional surface displacements maps by using the aforementioned technique. We show the main features of the approach proposed, and we discuss the limitations and potential future development of such optimization techniques for the analysis and interpretation of landslide phenomena and other geo-hazards.  

References

D. Albiol, R. Iglesias, F. Sánchez, and J. Duro, Improved characterization of slow-moving landslides by means of adaptive NL-InSAR filtering in Proc. SPIE: SAR Image Analysis, Modeling, and Techniques XIV, Amsterdam (Netherlands), Oct. 2014.

Manconi, A., P. Tizzani, S.Pepe, G. Zeni and G. Solaro, “Simulated Annealing and Genetic Algorithm Optimization Using COMSOL Multiphysics: Applications to the Analysis of Ground Deformation in Active Volcanic Areas”. In Proceedings of COMSOL European Conference 2009, Milan, Italy.

Tizzani, P., A. Manconi, G. Zeni, A. Pepe, M. Manzo, A. Camacho, and J. Fernández (2010), “Long-term versus short-term deformation processes at Tenerife (Canary Islands)”, J. Geophys. Res., 115, B12412, doi: 10.1029/2010JB007735.

The Slumgullion landslide in southwestern Colorado, USA is a natural laboratory for studying the mechanics of landslide motion because of its rapid motion of up to 2 cm/day and its large spatial extent. We are using interferometric synthetic aperture radar (InSAR) analysis of COSMO-SkyMed™ (CSK) X-band satellite Spotlight images and UAVSAR L-band airborne images to examine the distribution of motion within the landslide. The ultimate goal is to improve the understanding of the physical mechanisms controlling landslide motion by studying the landslide-wide kinematics of the Slumgullion landslide. Our earlier analysis of single CSK and UAVSAR interferograms confirmed the ground measurements of the changes in the landslide velocity over different seasons. We then used interferograms from the four different flight tracks of UAVSAR over Slumgullion to extract the full three-dimensional motion vector for every location in the landslide. Because of the rapid motion, we scheduled UAVSAR flights about a week apart to keep the total line-of-sight displacement to less than one L-band fringe in the interferograms. Then we combine the interferograms from the four viewing geometries to determine the three-dimensional motion over the weeklong time interval. We are now working on measuring the larger motion over longer time intervals of months or years using pixel offset tracking on the CSK and UAVSAR images. Both UAVSAR and CSK Spotlight mode have very high spatial resolution, so they can measure the Slumgullion landslide motion with pixel offsets over intervals as short as a month, and can easily image the motion over a year when the landslide has moved many meters. Because the pixel offset tracking measures two components of motion from each pair, we can extract the full three-dimensional motion from the pixel offsets from two different satellite tracks. The NASA UAVSAR flights include all four lines in a short time, which makes the combination simple. Combining the pixel offsets from the CSK satellites requires time series analysis to merge the scenes acquired on different dates.

Differential Synthetic Aperture Radar (SAR) Interferometry (DInSAR) represents a microwave remote sensing technique that allows retrieving surface displacements with centimetre to millimetre accuracy [1]. DInSAR technique has already well demonstrated its capability to detect and map landslides deformation phenomena, significantly contributing to advance the knowledge and the analyses of mass movement phenomena. The new challenge that the research in such a field is going to tackle is the effective integration and joint exploitation of remote sensing measurements with in situ data and observations, to study and understand the sub-surface interactions, the triggering causes, and, in general, the long term behaviour of the investigated landslide phenomenon. In this context, a very promising approach is represented by Finite Element (FE) techniques, which allow us to consider the intrinsic complexity of the mass movement phenomena and to effectively benefit from multi source observations and data [2].

In this study, we perform a three dimensional (3D) numerical model of the Ivancich (Assisi, Central Italy) instability phenomenon. In particular, we apply an inverse FE method based on a Genetic Algorithm optimization procedure [2], benefitting from advanced DInSAR measurements (Figure 1), retrieved through the full resolution Small Baseline Subset (SBAS) technique [3], and an inclinometric array distribution. To this purpose we consider the SAR images acquired from descending orbit by the COSMO-SkyMed (CSK) X-band radar constellation, from December 2009 to February 2012. The CSK data set is characterized by a reduced revisit time and improved spatial resolution with respect to the ERS-ENVISAT satellites, resulting in a 15-time larger point density [2]. Moreover the optimization input dataset is completed by an array of eleven inclinometer measurements, from 1999 to 2006, distributed along the unstable mass.

The landslide body is formed of debris material sliding on a arenaceous marl substratum, with a thin shear band detected using borehole and inclinometric data, at depth ranging from 20 to 60 m. Specifically, we consider the active role of this shear band in the control of the landslide evolution process. A large field monitoring dataset of the landslide process, including at-depth piezometric and geological borehole observations, were available. The integration of these datasets allows us to develop a 3D structural geological model of the considered slope (Figure 2).

Furthermore, the FE model of the active ground deformation field is performed in a fluid-dynamic physical scenario by considering different values of Newtonian viscosity in the shear band region. The model parameters (viscosities) are determined through an optimization procedure based on the minimization of the difference between the displacements retrieved by the model and the input datasets, represented by DInSAR and inclinometric displacements measurements.

The first experimental results are very promising and clearly show that this numerical approach is able to simulate the spatial distribution and temporal evolution of the landslide phenomenon kinematic, highlighting that the translational movement of such a landslide portion results to be the reason for the long-term instability. We conclude that the developed three dimensional physical model, built up by the integration of dataset derived from geological information and different measure platforms, represents an innovative and effective approach to improve the comprehension of the slow landslide processes.

References

[1]  R. Goldstein, and H.A. Zebker, “Mappings mall elevation changes over large areas: Differentia radar interferometry,” J. Geophys. Res., vol.94, no.B7, pp.9183–9191, 1989.

[2]  F. Calò, F. Ardizzone, R. Castaldo, P. Lollino, P. Tizzani, F. Guzzetti, R. Lanari, M.-G. Angeli, F. Pontoni, and M. Manunta, “Enhanced landslide investigations through advanced DInSAR techniques: The Ivancich case study, Assisi, Italy,” Remote Sens. Environ., vol. 142, pp. 69–82, Feb. 2014.

[3]  R. Lanari, O. Mora, M. Manunta, J. J. Mallorquí, P. Berardino, and E. Sansosti, “A Small-Baseline Approach for Investigating Deformations on Full-Resolution Differential SAR Interferograms,” IEEE Trans. Geosci. Remote Sens., vol. 42, no. 7, pp. 1377–1386, 2004.

It is established that InSAR is a very valuable tool for surface deformation mapping and monitoring for a wide area of applications. In Norway, we have during the last 10 years operationally applied InSAR data for landslide mapping and monitoring.

Over the last decade we have acquired several thousand ERS, Envisat ASAR, RADARSAT-2, and TerraSAR-X scenes, with variable spatial and temporal resolution. The data has been processed using both a small-baseline (SBAS) and a persistent scatterer (PSI) time-series methodology. Processing has revealed numerous previously undetected landslides, as well as documented many other ongoing periglacial processes such as solifluction
sheets and lobes, rock glaciers, blockfields and avalanche
fans.

The interferometric coherence is generally very good since most rockslides in northern Norway are located above 500–700 m elevation, where the vegetation cover is limited. However, the high latitude leads to long winters with snow cover main parts of the year. We can only include SAR images from the snow-free period (approximately June–September) in the InSAR processing, leading to a highly irregular sampling of the SAR data. With a revisit time of 35 days for ERS and ASAR, very few SAR scenes are available each summer. This limits the analysis to studies of long-term displacement trends. Intraseasonal displacement variations are difficult, if not impossible, to detect. Newer sensors such as TerraSAR-X with a shorter revisit time of 11 days allow for more scenes to be acquired during the snow-free season. However, the shorter wavelength of X-band makes it more challenging to detect long-term displacement rates, due to phase ambiguities between the different years.

Here, we discuss possibilities and limitations of InSAR as a tool for monitoring surface deformation in areas with annual freeze-thaw cycles. We demonstrate that by careful processing and interpretation, InSAR is able to pick up the seasonal deformation patterns of frost heave and ground settlement. We show examples using different sensors ranging from relatively low spatial resolution (ERS, Envisat ASAR) to very high resolution Staring Spotlight (TerraSAR-X), using different temporal sampling and radar frequencies. Finally, we show the first examples on using Sentinel-1 for monitoring surface processes in the arctic.

Round Table