PS-INSAR Monitoring of Landslides in the Great Caucasus, Russia, Using ENVISAT, ALOS and TerraSAR-X SAR Images
Mikhailov, Valentin1; Kiseleva, Elena1; Smolyaninova, Ekaterina1; Golubev, Vasily1; Dmitriev, Pavel1; Isaev, Yury2; Dorokhin, Kirill2; Hooper, Andy3; Samiei-Esfahany, Sami3; Hanssen, Ramon3
1Institute of physics of the Earth RAS, Moscow, RUSSIAN FEDERATION; 2“Lenmetrogiprotrans”, St. Petersburg, RUSSIAN FEDERATION; 3Delft University of Technology,, NETHERLANDS

We investigated two landslides at the foothills of the Northern Caucasus: (1) in the Mamaika village near the Black Sea beach above the tunnel of the Tuapse-Sochi railroad and (2) near the Kepsha village in the vicinity of the new railroad and highway from the Sochi airport up to the Sochi 2014 Olympic Games facilities in Krasnaya Polyana. The conditions for application of SAR interferometry for these two landslides are very different. The Mamaika landslide is located in highly populated area with lots of facilities being good reflectors while the Kepsha landslide is fully covered by vegetation and few reflectors can be revealed. The topography in the Mamaika landslide area is rather smooth: heights vary from 0 to 300 m. On the contrary, topography in the vicinity of the Kepsha landslide is very much dissected, heights range from 200 to 1200 m. It is worth noting that landslides are numerous in the North Caucasus region and satellite monitoring could help considerably in landslide risk assessment.
For PS-InSAR processing we used the StaMPS software (Stanford Method of Persistent Scatters). For Kepsha we incorporated 18 L-band ALOS images for the period 21.01.2007 - 17.09.2010, C-band Envisat images (from track 85A 11 acquisitions spanning period 01.08.2004-01.02.2009; from track 35D 12 acquisitions for the period 01.2007-09.2010 and 13 acquisitions for the period 29.11.2010 - 23.03.2012, i.e. before and after orbit manoeuvre) and X-band TerraSAR-X images (23 images, track 107D, 13.07.2012 - 07.02.2013). For Mamaika we used the same images from Envisat track 35D and also 23 images from TSX track 54A (24.12.2011 - 3.02.2013).
As it was expected the number of PS identified for the Mamaika landslide is much bigger than that for Kepsha. However even for Kepsha we also managed to identify sufficient quantity of persistent scatters to locate zones of comparatively high displacements. Success of PS-INSAR monitoring of surface displacements in mountainous and densely vegetated areas strongly depends upon strategy of processing. We describe our approaches to key processing procedures in particular, choice of "master" image, image cropping, selection of reference area and DEM. This helps to reduce influence of atmosphere and topo errors.
For the mostly urban area of the Mamaika landslide the best results were obtained using TSX data, while for the rural Kepsha landslide joint analysis of the data from different satellites with different wavelengths is preferable. Distribution of PS and corresponding time series for the latest sets of TSX images for the Mamaika landslide and Envisat images for the Kepsha landslide are shown in Figures 1 and 2.
Authors acknowledge the European Space Agency ESA (project C1-7991), the Japanese Space Agency JAXA and Deutsches Zentrum für Luft- und Raumfahrt DLR (project LAN1247) who kindly supplied us with SAR data for this study.

Fig.1 PS at the Mamaika landslide found from 23 TSX images spanning period 24.12.2011-07.11.2012. Figure at the top shows PS distribution and Vlos. PS's shown in red move towards the satellite, blue - away from it. Green PS have close to zero velocity. Color strength correlates with absolute value of Vlos. Figure at the bottom shows time series for test areas marked by yellow circles and numbers on the top figure.

Figure 2. PS’s on the Kepsha landslide found from 13 Envisat images spanning period 29.11 2010 - 23.03.2012 (after orbit transfer manoeuvre in October 2010). At the top is PS distribution and Vlos (table to the right). Numbers of PS in the table (in bold) correspond to the numbers of PS in the figure to the left. Positive Vlos directed towards the satellite. At the bottom are time series for PS's which Vlos is more than 3 mm/year. Numbers are the same as on the top figure and in the table.