Science Service System

Summary of Proposal LAN2548

TitleMapping creeping landforms in Alpine periglacial environment using TSX archives (Western Swiss Alps, Switzerland) – LAN1145 Extension
Investigator Barboux, Chloé - University of Fribourg, Department of Geosciences, Geography
Team Member
Prof. Delaloye, Reynald - University of Fribourg, Department of Geosciences, Geography
Prof. Collet, Claude - University of Fribourg, Department of Geosciences, Geography
Dr. Lambiel, Christophe - University of Lausanne, Institute of Earth Surface Dynamics
Dr. Strozzi, Tazio - Gamma Remote Sensing, N/A
Eng. Barboux, Chloé - University of Fribourg, Department of Geosciences, Geography

The project aims at both updating and upgrading the creeping landform inventory in the Western Swiss Alps using TSX stripmap mode interferometric data (single polarization, 3m resolution, 30 km swath width, 11 days acquisition time interval). It concentrates on areas situated above the tree line, which is mainly located in the Alpine periglacial belt. The existing inventory is essentially based on manual interpretation of a large set of C- and L-band DInSAR data from the 1990s and needs to be renewed (1,2,3).

Specific objectives of the proposed activity include:
- The update (changes in activity rate) and the upgrade (more accurate identification of active landforms and quantification of displacement rate) of the whole creeping landform inventory and especially active rock glaciers in the region of interest using TSX archives from 2008 to today (archives from LAN411and LAN1145 as well as archives requested in this proposal from orbits mentioned in the part Data requirements).
- The investigation of automated methods to detect creeping landforms in periglacial environment using a complete set of repeated TSX acquisitions during summer (archives requested in this proposal from orbits mentioned in the part “Data requirements”).

The previous studies have shown that a large set of SAR scenes covering several years and various time intervals was necessary to establish inventories of slope motion in a confident way (1-4). It is thus expected that the use of a TSX dataset as large as possible will surely increase the relevance of these existing inventories by allowing a more accurate detection of moving landforms and a better quantification of their displacement rate in many cases. Attempts to develop automated methods for the detection of slope movements also show the need of large TSX dataset (5). Finally, we would like to suggest the potential use of this kind of complete set of repeated acquisitions for detecting a change (automatically?) in the deformation rate of active landforms moving with a velocity rate in the order of cm/month to dm/month.

References :
  1. Delaloye R., Lambiel C. & Lugon R. (2005). ESA SLAM project, phase 2, Bas-Valais. Validation of InSAR data in permafrost zone.Unpublished.
  2. Delaloye R., Perruchoud E., LambielC., Lugon R. (2008), InSAR Haut-Valais: Inventaire des mouvements de terrain par analyse de signaux d’interférométrie radar satellitaire (période 1993-2000), Rapport final, Mandant: Canton du Valais.
  3. Barboux C., Lambiel C., Delaloye R. (2014) Mapping of slope movements in Alpine environment using DInSAR. Earth Surface Processes and Landforms.(submitted)
  4. Barboux C., Delaloye R., Lambiel C., Strozzi T., Collet C. & RaetzoH. (2013). Surveying the activity of permafrost landforms in the Valais Alps with InSAR. Mattertal- ein Tal in Bewegung. Publikation zur Jahrestagung der Schweizerischen Geomorphologischen Gesellschaft 29. Juni - 1 Juli 2011, St Niklaus
  5. Barboux C., Delaloye R., Strozzi T., Lambiel C. & Collet C. (2013) TSX DInSAR data for detecting and monitoring phenomena in an alpine periglacial environment at different resolution scales (Western Swiss Alps, Switzerland). TerraSAR-X Science Team Meeting, 10-12 June 2013, DLR, Oberpfaffenhofen.
Detailed report

The innovative character of this project was to investigate automated techniques to detect and map slope movements in Alpine periglacial environment. The objective of this project was to - so far as possible - automatically update existing inventories of slope movements of Western Swiss Alps by integrating the most recent DInSAR data in order to detect potential change in activity rate of landforms (1). The current inventory indicates the outlines of moving zones through detected signals of different magnitude orders (cm/day, dm/month, cm/month, cm/year). It contains signal patterns that are related to different phenomena like glaciers, debris-covered glaciers, push-moraines, active rock glaciers, landslides and saggings. In order to obtain new information of active rock glaciers and moving landslides in the studied area, the plan was to use a large set of TSX SAR data (2008-2013).
To support these targets, a large set of TSX SAR data was used covering our region of interest from the projects LAN411, LAN1145 and LAN2458 together. DGPS measurements from campaigns and permanent GPS stations acquired over more than 20 sites during 2008-2013 have also been used to compare and validate the results.

This project proposes a semi-automated method to map the slope movements in the Alpine environment from TSX interferometric phase and coherence images. The resulting map of slope movements provides a general overview of the surface deformation occurring over the area during a specific time interval (2-4).
The protocol was developed in the basis of past studies using visual interpretation of DInSAR data for inventorying the Alpine slope movement in the Swiss Alps. Like the eyes of the expert, the process aims to classify the phase image texture in order to detect change of the phase image related to surface deformation (5, fig.1). Thus, the procedure uses segmentation and classification methodologies applied on interferometric coherence and phase images to detect and map each pixel according to one of the three defined patterns of DInSAR signal, namely: plain, (partly) fringe or noise pattern. Then, by combining maps of DInSAR signal from a set of selected interferogram having the same time lapse, the classification is performed in term of surface deformation and the identification is mapped according to one of the three defined class of deformation, namely 1) area without deformation, 2) with gentle deformation or 3) with large deformation (2-4).
The model of creating maps of slope movements is simple but robust and the classification of the slope movements is much more reliable by using a large number of DInSAR pairs. Different models were developed and the performance was evaluated according to the visual interpretation established by experts, DGPS measurements as well as proposition regarding DInSAR signal evolution (2-4). Finally, a standard procedure with specific parameters values was recommended to accurately map slope movements in our region of interest (2-4). These values are dependent of the observed moving landforms (mainly size and velocity) and parameters should be adapted if applied in another context.

Finally, when using a defined set of parameters, the proposed procedure performs automatically the detection and mapping of DInSAR signal.
However, the resulting map of slope movements needs expertise to be correctly interpreted. Indeed, the areas identified as large deformations (related to a noise pattern in term of DInSAR signal) have by definition to be taken into account carefully to determine if they are really related to rapid slope movement or external artefact. In mountainous area especially, the temporal decorrelation due to the change of the scattering geometry from wet snow or vegetation for instance, and/or due to the change in the dielectric constant of the ground in between the two satellite acquisitions can also cause noise in the resulting phase image. Whereas the noise due to exceptional snowfall affects generally few pairs of phase images, the noise induced by vegetation is always present on the same location in the image. Moreover, a localized smooth change of the signal phase characterized by a (partly) fringe pattern may indicate a change of the surface geometry that can be quantified. However, it could also be due to external artefacts. When mapping the slope movements, maps of DInSAR signal derived from several phase images having the same time interval are combined together. Pixels are classified into the most represented DInSAR signal class in the set of selected interferograms. Thus, the combination of DInSAR signal maps from several interferograms having the same time interval is a necessary and sufficient condition to prevent from single external artefacts. But for further investigations, the combined use of the slope movement maps with visual expert interpretations is absolutely recommended in order to produce a much more reliable analysis of slope movement in the region of interest, less subjective and fast.

Applications investigated using slope movement map

1) Rough estimation of deformation rate
The slope movement map derived from a set of DInSAR signal maps with the same time interval permits the identification of the related areas without deformation, with gentle deformation and with large deformation at this specific time interval. The rate of terrain movement which can be detected depends among others on the time interval and on the wavelength. The interferometric SAR signal will decorrelate when the displacement gradient between adjacent pixels is higher than half the wavelength during the selected time interval. Consequently, it is possible to roughly evaluate the deformation rate on a spatially outlined area with an almost homogeneous deformation rate by analyzing the evolution according to time interval of each proportion of slope movements given by the automated mapping (fig. 2). Two thresholds can be detected: 1) s_stab corresponding to the time interval until which the proportion of plain pattern is higher than 50% and 2) s_dec corresponding to the time interval from which the proportion of noise pattern is higher than 50%. The period in between these two thresholds represents the time intervals where (partly) fringe pattern can be observed. Finally, the deformation rate can finally be evaluated according to the 4 classes used in previous inventories of slope movements using DInSAR, namely the classes “cm/year”, “cm/month”, “dm/month” and “cm/day” (2-4).

2) Update of existing inventories The automatic rough estimation of deformation rate was used in a small studied area using large set of TSX DInSAR scenes from summers 2008 to 2012 in order to update past moving slope inventories (2-4, fig.3). The method was evaluated by analyzing the accuracy to detect a change in deformation rate of DInSAR polygons. False change detection is mainly due to external factors as vegetation, snow or atmosphere (where the signal is noisy), due to border effect in layover and shadow areas, as well as due to a change of the outline of the landform.
The compatibility between different sensors has also to be considered. The outline of the slope movement can basically differ due to different ground resolution, incidence angle or acquisition mode. In comparison to ERS, recent very high-resolution X-Band sensors as TerraSAR-X (TSX) permit a more precise spatial detection and delimitation of moving slopes. Moreover, with a repeated cycle of 11-days respectively, the moving slopes observed on interferograms subtle change of deformation rate can be detected.
By improving the spatio-temporal resolution capacity of detection, the spatio-temporal variability of the moving slopes geometry increases too. Thus, the comparison of the mapping of slope movements derived from different sensor technologies may be a difficult task and will have to be performed carefully.

3) Upgrade of existing inventories
The visual update and upgrade of existing DInSAR polygons over our region of interest, passing through the past ERS technology to the current TSX technology, seems to be the more accurate solution due to a problem of sensor compatibility (fig.4). Actually, the margins of detected moving zones and the correspondence between DInSAR signal patterns and deformation rates differ according to SAR sensor involving difficulties for an automated approach (2-4).
However, this kind of slope movement maps developed in this project may support expert in the development of accurate slope movement inventory free from the subjectivity of the operator by using it as a useful tool for visual interpretation of DInSAR data especially when using large SAR dataset.

4) Monitoring of slope movements
To go further, we also expect that the use of this kind of complete set of repeated images acquired during the summer allows the detection of seasonal variations and geomorphologic process of specific landforms. In other words, the 11 days regular acquisitions of TSX data could be used as a potential early warning system. To perform such of analysis, we propose to observe the evolution of the DInSAR signal over the selected landform during the summer season on a complete set of interferograms with a time interval of 11 days. We suggest that the variation of the proportion of each DInSAR signal pattern (plain, (partly) fringe and noise) is related to the variation of deformation and that this analysis allows the qualitative estimation of seasonal variation of the landform deformation rate (2, fig.5).

Finally, this project has shown that it is possible to develop automated methods for the detection and the mapping of slope movements in an Alpine environment. The proposed procedure delivers maps providing a general overview of the surface deformation occurring over the area during a specific time interval. These mapping of slope movements can be used to update automatically the existing inventories of DInSAR polygons. However, experiments show that manual upgrade is maybe the most reliable technique and that these maps are all the same useful to help expert as a visual tool for the interpretation of surface deformation and by reducing the subjectivity. We also suggest an additional use of these maps, much more local, allowing the survey of specific landforms in term of changing activity.

1. Delaloye R., Lambiel C., Lugon R., Raetzo H., Strozzi T. (2006). ERS InSAR for detecting slope movement in a periglacial mountain environment (western Valais Alps, Switzerland). High Mountain Proceedings of the 9th International Symposium on High Mountain Remote Sensing Cartography (HMRSC-IX), Graz, Austria, 14-15 Sept. 2006.
2. Barboux, C. (2015). Detection, mapping and monitoring of slope movements in the Alpine environment using DInSAR. PhD Thesis. Departement of Geosciences, Geography, University of Fribourg. (submitted)
3. Barboux, C., Delaloye, R., Strozzi, T., Lambiel, C., Raetzo, H. and Collet, C. (2013). Semi-automated detection of terrain stability in the Swiss Alpine periglacial environment using segmentation and classification of DInSAR scenes: a useful tool to update past inventories of moving areas. Living Planet Symposium. 09-13 september 2013 Edimburg, Scotland.
4. Barboux, C., Delaloye, R., Strozzi, T., Lambiel, C. and Collet, C. (2013). TSX DInSAR data for detecting and monitoring slope motion phenomena in an Alpine periglacial environment at different resolution scales (Western Swiss Alps, Switzerland) - LAN 1145. TerraSAR-X / TanDEM-X science team meeting, 10-14 June 2013, DLR Oberpfaffenhofen, Germany.
5 - Barboux, C., Delaloye, Lambiel, C. (2014). Inventorying slope movements in Alpine environment using DInSAR. Earth Surface Processes and Landforms. DOI: 10.1002/esp.3603 (In press).

Back to list of proposals

© DLR 2004-2016