and Tobias Bolch TU Dresden, Universität Zürich D5P2a Optical and Thermal 20/06/2013 1 20/06/2013 1 Outline 1. s 2. Spectral Properties of s 3. of s a. Data Sources b. Automated mapping c. Postprocessing 4. of Debris Covered s 5. Generation of a a. Splitting Contiguous Ice Masses b. Assigning Topographic Parameters 6. Estimation of 7. 8. 9. 10. 20/06/2013 2 20/06/2013 2 1
s Tschierva /Alps Why should we study glaciers? Foto: T. Bolch 20/06/2013 3 20/06/2013 3 s Landsat ASTER auf ETM auf auf 1992 1999 2004 SRTM/ASTER-DGM 20/06/2013 4 20/06/2013 4 2
s Cumulative Length from Field Measurements 20/06/2013 5 20/06/2013 5 s Why should we investigate glaciers? changes can be easily recognised even by a layman. Fluctuations of a glacier, which are not influenced by thick debris, calving or surge instabilities, are a reaction to climatic forcing. changes are recognised as high confident climate indicator and as a valuable element in early detection strategies (GCOS 2004, GTOS 2008), s exist in remote areas where climate stations are rare. length change is the indirect, delayed, filtered but also enhanced signal to a change in climate, mass balance (i.e., the change in thickness/volume) is the direct and un delayed response to the annual atmospheric conditions (WGMS 2008). 20/06/2013 6 20/06/2013 6 3
s Why should we investigate glaciers? s are important sources of fresh water especially in arid regions with low summer precipitation. melt can result in increasing lake levels flooding pastures in endorehic basins. melt water contributes significantly to global sealevel rise. Receding glaciers can cause serious hazards e.g. by destabilized slopes or outbursts of glacial lakes. s are attractive elements of the natural landscape and a tourist attraction. 20/06/2013 7 20/06/2013 7 s Accumulation Area + Ablation Area - Mass Gain Equilibrium Line (Mass Balance = 0) Mass Loss Warming Increase Equilibrium Line Cooling Lowering Thichness Change Length Change 20/06/2013 8 20/06/2013 8 8 4
s Tschierva /Alps Accumulation area Ablation area Equilibrium Line Debris on Lateral Moraines Foto: T. Bolch 20/06/2013 9 20/06/2013 9 Terminus 2003 s Glacial Lake Moraine of the Little Ice Age Foto: T. Bolch 2003 20/06/2013 10 20/06/2013 10 5
s Late Summer Snow Line ( Equilibrium Line) Foto: T. Bolch 20/06/2013 11 20/06/2013 11 Introduction One Outlet of Vatnajökull (Water ) in Iceland s Photo: T. Bolch 20/06/2013 12 20/06/2013 12 6
Introduction Debris covered glacier at Mt. Everest s Photo: T. Bolch 20/06/2013 13 20/06/2013 13 Rockglacier in Northern Tien Shan s Photo: T. Bolch 20/06/2013 14 20/06/2013 14 7
Introduction Eastern Pamir s Photo: T. Bolch 2005 20/06/2013 15 20/06/2013 15 Introduction Zhadang /Nyanchênthangla, Tibet s Foto: T. Pieczonka, Oct. 2009 20/06/2013 16 20/06/2013 16 8
Typical Zones of a s Cuffey and Paterson (2010) 20/06/2013 17 20/06/2013 17 Introduction Change s Important Measures for Length Area Mass Balance Equilibrium Line Altitude 20/06/2013 18 20/06/2013 18 9
s => The area showed a delayed signal Tsentralny Tuyuksu 1998 1970 Foto: Foto: K. G. V. Makarevich N. Uvarov Mass Budget Area Quelle: WGMS 20/06/2013 19 20/06/2013 19 Glaciated Area ~1960 1999: - 16 % s => Length and area changes can be influenced by non climatic factors 20/06/2013 20 20/06/2013 20 10
Data s Data and Imagery for Studies ~1900 Historical Maps and oblique Photos ~1940 Aerial Imagery (pan) 1960 1980 Spy Sat. (e.g. Corona, <8m, pan, stereo) 1972 Landsat MSS (80m, VNIR), KFA 1000 (5m, pan) 1982 Landsat TM (30m), 1999 ETM+ (15/30m) 2000 Ikonos, Quickbird Ultra high resolution <1 4 m 2000 ASTER (15m, stereo) 2007 ALOS (2.5m, stereo) Others: MK 1000, Spot, IRS, Cartosat, CBERS 2etc. 20/06/2013 21 20/06/2013 21 Sensors Suitable for Automated s 20/06/2013 22 20/06/2013 22 11
Data Historical Maps Mapa (~1900 1937) s Map of Zemu from German Expedition (Finsterwalder 1935) 20/06/2013 23 20/06/2013 23 Data Oblique Photos s Robson (Rocky Mountains/Canada) 1908 2004 20/06/2013 24 20/06/2013 24 12
Data Aerial Images s Northern Tien Shan 1956 20/06/2013 25 20/06/2013 25 Data s First space borne images: US Spy Images Corona (since 1960) Corona KH 4 (since 1962) Corona KH 4B (since 1972) Panchromatic images Spatial Resolution: 8 2 (!) m Footprint: ~15x220 km 20/06/2013 26 20/06/2013 26 13
Data First continued global mapping: Landsat MSS (since 1972) s 3 visible and 1 near infrared band Spatial resolution: 79m Footprint: 170 x 180km 20/06/2013 27 20/06/2013 27 Data s Hexagon KH 9 (1963 1980) 120x120 km Resolution: 6 10m 20/06/2013 28 20/06/2013 28 14
Data s First Satellite images with short wave infrared band: Landsat TM (since 1982) Spatial Resolution: 30m Footprint: 170x180km Further important sensors: SPOT (since 1986) PAN/VN /SWIR, 10/20m, 60x60 km IRS ASTER (since 1999) VN /SWIR, 15m, 60x60km 20/06/2013 29 20/06/2013 29 Data Landsat ETM+ (since 1999) Scan line failure since 2003 (SLCoff scenes) 20/06/2013 30 20/06/2013 30 15
Very high resolution images Ikonos/Quickbird (since 2000) s 20/06/2013 31 20/06/2013 31 s Ikonos (2000) 1962 (Corona) 2003 (ASTER) 20/06/2013 32 20/06/2013 32 16
s Remote Sening Imagery can be used to a. Extract glacier extent b. Identify surface characteristics c. Determine volume loss d. Calcualate glacier velocity These characteristcs can be obtained Manually (time consuming, but more accurate?) Automated(faster, accurate but only for clean ice) 20/06/2013 33 20/06/2013 33 s ρa= E R (λ) / E I (λ) whereas ρa= spectral reflectance E R (λ) = energy of the wavelength reflected from the object E I (λ) = energy of the wavelength received by the object Fresh Snow Paul, 2003 Dirty Ice Hall et al., 1985 Ice 20/06/2013 34 20/06/2013 34 17
s Wave Length Bolch 2006 after Kääb et al. 20/06/2013 35 20/06/2013 35 Remote Sensing of s s TM 3: Red (0.63 0.69µm) TM 5: SWIR (1.55 1.75µm) TM 4: NIR (0.76 0.90µm) TM 6: TIR (10.4 12.5µm) 20/06/2013 36 20/06/2013 36 18
Remote Sensing of s Influence of Atmosphere Influence of Topography s No corrections are required for glaciers mapping based on image ratioing 20/06/2013 37 20/06/2013 37 s Exercise 1: Visualizing multi spectral images with Erdas Imagine 2011 View raster data: Open file: File > Open > Raster Layer; select: E:/Data/DP5P2a/l7_20050710_30m.img The provided data is a subset of an orthorectified Landsat 7 ETM+ scene contains all layers in a single file. The scene is acquired 7 Oct. 2005 and represents the central part of western Nyainqentanglha range in south central Tibet. The stripes in the upper left are due to an scan line failure of the instrument since May 2003. View the metadata of an image: Right click with the mouse on the file name in the table of contents or Home > Metadata Question: What is the pixel size and the projection? 20/06/2013 38 20/06/2013 38 19
s Exercise 1: Visualizing multi spectral images with Erdas Imagine 2011 Visualizing an image in true and false colours First improve the contrast of the image (Image enhancement): Multispectral > General contrast > different options (try Standard Deviation Stretch ) Choose the different bands for displaying in the RBG mode: Multispectral > Bands > Choose the different bands in the drop down menu Try different band combinations (e.g. 3 2 1, 4 3 2 and 5 4 3) You can see one specific band when selecting the same band for RGB. Question: Which is best band combination for visualizing the glaciers and why? 20/06/2013 39 20/06/2013 39 s Exercise 1: Visualizing multi spectral images with Erdas Imagine 2011 Zoom in and out, pan etc. Several possibilities to zoom etc. can be found In the toolbar of Home represented the typical icons. Check values of the pixels The pixel values range from 0 until 255 (8 bit) The value of each single pixel can be seen as follows: Home > Inquire >Inquire > A cross appears over the image and a window opens in which the pixel value of the file and the display can be seen. You can move the cross to each pixel to see its value(s). 20/06/2013 40 20/06/2013 40 20
Main Workflow for Generation of s F. Paul/s_cci 20/06/2013 41 20/06/2013 41 s Processing: First crucial step: Selection of a suitable scene At the end of the ablation period No seasonal snow, no clouds Small shadows => Best time varies according to location and climatic zone but is usually end of August Landsat is most often the first choice: free, suitable resolution, large footprint, and orthorectified; best search site: http://glovis.usgs.gov 20/06/2013 42 20/06/2013 42 21
Exercise 2: Searching for suitable satellite data We use the following region in south central Tibet as an example: s Figure 1: Map of western Nyainqentanglha (source: Bolch et al. 2010, TC) 20/06/2013 43 20/06/2013 43 s Exercise 2: Searching for suitable satellite data Landsat and other data can be downloaded for free from USGS Global Visualization Viewer http://glovis.usgs.gov Select the best scenes Suggested settings: Collection > Landsat Archive > L4 7 combined Resolution: 240m Maximum cloud cover: 30% Provide either the path/row of the Landsat scene or the center coordinate of the region of interest: Latitude: 30.3, Longitude: 90.6 Select date of interest: Start with June 2000 (usually first suitable data for Landsat ETM+ data) Go backwards in time and then forward. What do you realize about the availability of the scenes? Search the most suitable scene for the detailed study region (see. Fig. 1). Question: Which scene would you use for the glacier inventory and why? 20/06/2013 44 20/06/2013 44 22
s Exercise 2: Searching for suitable satellite data In case the GLOVIS website is not running (java required ): Use www.landcover.org Klick on ESDI (upper right below Download Data ) Map Search Enter the coordinated from above (same for Min and Max) Preview and Download Select the most suitable scenes (click on ID) form Ortho, Geocover (best orthorectified product). You can enlarge the scene when clicking on the thumbnail image Question: What is the most suitable scene and why? 20/06/2013 45 20/06/2013 45 Main Workflow for Generation of s F. Paul/s_cci 20/06/2013 46 20/06/2013 46 23
s Processing: Supervised Classification Ratio Image Segmentation (thresholding) Ratio NIR/SWIR (e.g. TM 4 / 5, snow/ice >~2.0) Also RED/SWIR (3/5, snow/ice > ~2.0, 'better' for shadow areas, and snow versus ice separation) Other Sensor: ASTER 3/4 Normalised Difference Snow Index (NDSI): (VIS SWIR)/(VIS+SWIR), e.g. (2 5)/(2+5) or (3 5/3+5) 20/06/2013 47 20/06/2013 47 Remote Sensing of s Delineation Supervised Classification s 20/06/2013 48 20/06/2013 48 24
Remote Sensing of s Delineation s White: Snow Light blue: Firn Dark blue: Ice Red: Debris Supervised Classification Sidjak & Wheate (1999) 20/06/2013 49 20/06/2013 49 s 20/06/2013 50 20/06/2013 50 25
TM3/TM5Ratio Image Image Ratioing TM4/TM5 > 2 (white) TM3/TM5 > 2 (+ red) s Critical Step: Identification of best threshold => Visual checking is required Problem: Water bodies are also mapped NDWI and NDVI -Thresholding for lakes and vegetated area 20/06/2013 20/06/2013 white = blue = Lakes green = Vegetated Area 51 Bolch and Kamp (2006) Quelle: Bolch 2006, Diss 51 s 20/06/2013 Improved Landsat Ratio Water TRIM Raw 3/5 Ice Detection TM >= Layer Outlines 7-4-2 2 as (NDWI) Outlines Mask 20/06/2013 52 52 26
s Exercise 3: of s using Multi Spectral Images Method: Ratio RED/SWIR (TM3/TM5) Realization in ERDAS: Computing the ratio image A simple model needs to be build with the Model Maker I prepared a simple model for the calculation (band_ratio.gmd). Workflow: Toolbox > Model Maker > File > Open > Select band_ratio.gmd A graphical view of the simple model is opening. Double click on the symbol below Input raster and select l7_20050710_30m.img. The default setting needs not to be altered. Click OK Double click on the output raster and provide a file name l7_20050710_ratio35_raw.img Select data type float single if not already set by default. Click OK. Execute the model: process > run (or click on the red flash) Add now the result to the 2D view (file > open raster layer) 20/06/2013 53 20/06/2013 53 s Exercise 3: of s using Multi Spectral Images Search for the best threshold First improve the contrast, if required (panchromatic > general contrast) Check with the inquire cross the best threshold of the ratio image, the value should be around 2.0. Check especially the pixels around the glacier margins. You can blend between the two images with Home >Swipe or Blend Apply the threshold (begin.) We need to build a simple model for the calculation: First open the model builder. Move two icons raster layer (second highest left) and the circle icon into the empty space. Connect the icons with the arrow (from Raster to Circle to Raster ): Click on the arrow, move to the first object and draw the line from the first to the second object. 20/06/2013 54 20/06/2013 54 27
s Exercise 3: of s using Multi Spectral Images Apply the threshold (cont.) Double click on the first raster icon and choose the ratio image as input. Double click on the second raster and define a suitable name for the output (e.g. l7_20050710_glacier_raw.img ). The output data type should be thematical. Double click on the circle > a window opens; choose the Relational function from the drop down menu (not really needed for this operation as one can also type it in with the keyboard). Double click now on the correct input file which should be greater or equal (>=) the selected threshold. Type the selected threshold (e.g. 2.1). click OK Run the model. The output file is now the raw glacier mask. Analyse the raw result: What are the problematic areas? 20/06/2013 55 20/06/2013 55 Problem: Debris Cover s Shadow Postprocessing: Automated: Filter to eliminate isolated pixles Finally: Manual ajustment Clouds 20/06/2013 56 20/06/2013 56 28
s Bolch & Kamp 2006 20/06/2013 57 20/06/2013 57 s Exercise 3: of s using Multi Spectral Images Automated postprocessing In a postprocessing step we remove now some noise and isolated pixels automatically. There a different opportunities, such as smoothing. Very suitable is applying a statistical filter: Raster > Resolution > Spatial Statistical Filter, multiplier e.g. 2.0 > name output file (e.g. l7_20050107_glac_filt.img ) > OK Raster Vector Conversion A vector layer is required for further analysis Vector > Raster to ArcCoverage Rem.: The coverage is an older ESRI format and consists of a point, line and polygon file and an info table. The area is calculated automatically. Visually analyse the results: What parts of the glaciers are not mapped and why? 20/06/2013 58 20/06/2013 58 29
Exercise 3: of s using Multi Spectral Images Postprocessing in the vector domain Delete values with grid code = 0 Realization: Highlight vector layer Table > Criteria > grid code == 0 > select Drawing > delete selected element (red X) Delete misclassified pixels (e.g. small isolated snow patches, water) Automated using a size threshold (common 0.02 0.05 km²) Table > Criteria > area < 50,000 > select Drawing > delete selected element (red X) Manual select and delete remaining obviously misclassified polygon Save layer: right click > save layer 20/06/2013 59 20/06/2013 59 Debris Covered s (Example: Mt. Everest Region) s 20/06/2013 60 20/06/2013 60 30
s Slope Gradient Curvature(s) 20/06/2013 61 20/06/2013 61 Delineation of Debris Covered s Approaches s Thermal Information 20/06/2013 62 20/06/2013 62 31
s Slope Gradient Blue < 12 Red < 24 Delineation of Debris Covered s 20/06/2013 63 20/06/2013 63 s Supervised Spectral Classification and Slope Gradient (<12 ) Discrepancy to visually mapped Outlines: ~ 5% Approaches Delineation of Debris Covered s Bolch et al. (2007) 20/06/2013 64 20/06/2013 64 32
Approach to Map Debris covered s s Postprocessing: Filling Small Gaps Bolch et al. (2007) Bhambri et al. (2011) 20/06/2013 65 20/06/2013 65 Delineation of Debris Covered s Approaches Results s Deviation from Reference ~5% Main problem: correct delineation of front parts of glacier tongues Bolch et al. (2007) 20/06/2013 66 20/06/2013 66 33
with SAR Coherence Data s Atwood et al. (2010), CJRS 20/06/2013 67 20/06/2013 67 with SAR Coherence Data s Frey et al. 2012 (RSE) => Manual improvement (with additional information, if available) is still the most accurate method. 20/06/2013 68 20/06/2013 68 34
s Exercise 4: Debris covered s The main problem for the delineation of debris covered glaciers is the correct identification of the terminus of the glacier. The most accurate method is still the manual interpretation. There is one characteristic debris covered glacier in our test site (Question: Why is the glacier debris covered?): Xibu s (for location see Fig. 1). This glacier shall now be manually delineated and its area change between 1976 and 2005 investigated. Realization in ERDAS: Open file: File > Open > Raster Layer; select: E:/Data/DP5P2a/l7_20050710_15m.img l7_2 and kh9_1974.img Drawing a polygon: Vector > Drawing > Insert Geometry > Click on Polygon symbol 20/06/2013 69 20/06/2013 69 Main Workflow for a s F. Paul/s_cci 20/06/2013 70 20/06/2013 70 35
s Splitting the Polygons into their Drainage Basins Additional data: Co Registered Digital Elevation Model Manual digitization using Shaded relief Contour lines direction grid, and Satellite image as baseline information. Automated: based on hydrological analysis 20/06/2013 71 20/06/2013 71 s Splitting the Polygons into their Drainage Basins Simple automated Algorithm: 20/06/2013 72 20/06/2013 72 36
Splitting the Polygons into their Drainage Basins s Bolch et al. (2010). RSE 20/06/2013 73 20/06/2013 73 s Def.: map (outlines) plus topographic information Required minimum information: type Area Elevation (minimum, maximum, mean, median) Aspect Mean Slope Date of image acquisition Additional possible information: Percentage debris cover Late summer snow line Aspect Accumulation and ablation area Calculation based on DEM and GIS software (not available here) 20/06/2013 74 20/06/2013 74 37
s ±?? 20/06/2013 75 20/06/2013 75 Identification of sources causing uncertainty Main Error Sources Geolocation wrong threshold selected wrong or missed correction wrong position Type of Error Random Systematic Random Random poor DEM Random match (date, resolution) 20/06/2013 76 20/06/2013 76 38
s Systematic error: Area is under or overestimated when the threshold is too high / low mapped area too large too small 20/06/2013 77 20/06/2013 77 Random error / systematic error due to debris on ice Area : 6.82 km² Area Photo: 7.90 km² Deviation: 1.08 km² (+13.7%) s 20/06/2013 78 20/06/2013 78 39
Accuracy Assessment Random Error s Internal Rocks Seasonal snow Ice divide Area : 9.90 km² Area Photo: 9.66 km² Deviation: 0.24 km² (2.4%) 20/06/2013 79 20/06/2013 79 Random error: Outlines are wrongly interpreted due to debris cover or bare rock Landsat 30 m automatic with Landsat TM Ikonos 1 m terminus? debris Aerial 1 m bare rock 20/06/2013 debris 80 20/06/2013 80 80 40
Random error: different DEMs result in different locations of drainge divides s 20/06/2013 81 LeBris et al. (2011) 19/5/2011 20/06/2013 s CCI Meeting Thomas Nagler 81 81 Random error: wrong topographic parameters are derived (here min. elevation) when the DEM acquisition date does not match to the satellite data s 20/06/2013 82 19/5/2011 20/06/2013 s CCI Meeting 82 82 41
(Area): Multi temporal Scenes 20/06/2013 83 20/06/2013 83 : Multi temporal Scenes s 20/06/2013 84 20/06/2013 84 42
Hexagon/MSS Data (1976) s 20/06/2013 85 20/06/2013 85 s ( / Mass changes) length change is the indirect, delayed, filtered but also enhanced signal to a change in climate, mass balance (i.e., the change in thickness/volume) is the direct and un delayed response to the annual atmospheric conditions (WGMS 2008). changes can be calculated based on multi temporal Mass Gain DEMs Equilibrium Line (Mass Balance = 0) Mass Loss Thickness Change 20/06/2013 86 20/06/2013 86 43
1962 (Corona KH4) 1972 (Corona KH4B) 1984 (Wild RC10) 2001 (ASTER) 2007 (Cartosat-1) 20/06/2013 87 20/06/2013 87 2 Methods and Data Digital Elevation Models Improved ASTER DEM s Buchroithner & Bolch (2007) 20/06/2013 88 20/06/2013 88 44
2 Methods and Data Digital Elevation Models s DEM Comparison Original DEMs, e.g. Cartosat KH4B -> Tilt 20/06/2013 89 20/06/2013 89 2 Methods and Data Digital Elevation Models DEM Comparison, Original DEMs, e.g. Cartosat KH4B, deviation in absolute elevation, shift -> adjustment necessary s Count Deviation (m) Bolch et al. 2008, J. Glac. Pieczonka et al. 2011, ISPRS 20/06/2013 90 20/06/2013 90 45
s Systematic xy shift 20/06/2013 91 20/06/2013 91 Correction for a Systematic shift s Nuth and Kääb, (2011) 20/06/2013 92 20/06/2013 92 46
Mt. Everest Region s DEM Comparison Cartosat KH4B, (2007 1972) adjusted Surface lowering -0.36 ± 0.09 m/a Mass balance: -0.32 ± 0.08 m w.e./a For conversion into mass changes density assumptions are requited (ice density 900 kg/m³) Bolch et al. (2011) Piezonka et al. (2011) 20/06/2013 93 20/06/2013 93 RADAR/Laser Altimetry for estimation of elevation changes Radar Altimetry (Radar = Radio Detection and Ranging) pulsed micro waves are emitted and detected either air or space borne s Elevation results from the time between the emission and detection of the reflected pulse. The time is calculated based on the number of waves. Laser Altimetry is based on the same principle (puled light instead of radar data (Lidar [Light Detection and Ranging]). 20/06/2013 94 20/06/2013 94 47
ICESat tracks do not repeat exactly: Corrections must be applied s Moholdt et al. (2010) 20/06/2013 95 20/06/2013 95 ICESat GLAS analysis over Greenland s Pritchard et al. (2009) Bolch et al. (in rev.) 96 20/06/2013 96 20/06/2013 96 48
Remote Sensing of s Velocity s ASTER 3-3-1 over ASTER DEM; Visualisation: T. Bolch 20/06/2013 20/06/2013 97 97 Remote Sensing of s Velocity s Scud (2002) 20/06/2013 20/06/2013 98 98 49
Remote Sensing of s Velocity s Scud (2003) 0.5 km 20/06/2013 99 20/06/2013 99 s There are two main possibilities to derive the glacier flow at its surface from remote sensing data: Tracking of features on the glaciers surface (Feature tracking) Differential interferometric SAR (DInSAR) 20/06/2013 100 20/06/2013 100 50
Velocity s Cross Correlation Technique Volmer & Kääb (2000) Kääb (2006) 20/06/2013 101 20/06/2013 101 Remote Sensing of s Velocity velocity: (Examples: ASTER and Ikonos) s Bolch et al. (2008), NHESS 20/06/2013 102 20/06/2013 102 51
Khumbu Himalaya Velocity Image source: Ikonos 2000 & 2001 s Estimated error: ±7.7 m End of active glacier tongue 20/06/2013 103 20/06/2013 103 Bolch et al. (2008) s Velocity of Greenland ice sheet derived by micro wave data 20/06/2013 104 20/06/2013 104 Moon et al. 2012, Science 52
Exercise 6: Area / Search and Download suitable imagery from different time (here: use scene l7_20010502_30.img, from 02.05.2001) Classification (see exercise 3) Postprocessing (use also existing outlines as additional information) Split contiguous ice masses into single glaciers (see exercise 5, not possible here) Assign same id for glaciers as for those from the other period (not possible here) The change of the debris covered glaciers has to be manually adjusted (see exercise 4) Change Analysis (visual and by value) Sizes are displayed in the info table (highlight the respective file, then click on table ) Are the changes significant? What are the problems? 20/06/2013 105 20/06/2013 105 Thank you for your interest! Tobias Bolch tobias.bolch@geo.uzh.ch 20/06/2013 106 20/06/2013 106 53