On the use of Pléiades imagery in glaciology. DEM generation and velocity mapping Etienne Berthier (LEGOS) with colleagues from Univ. Iceland / IANIGLA-Mendoza / LGGE-LTHE-France Pléiades 3D view of the Tungnafellsjökull (Iceland). © CNES 2013, Distribution Airbus D&S Scientific context. Glaciers as indicators of climate change, contributor to sea level rise and water resources Hausse observée +3.2 mm/a Dilatation thermique +1.0 mm/a Observed SLR ~3.2 0.4 Calottes Polaires +0.6 mm/a Ocean warming +1.1 0.3 Glaciers +0.9 mm/a Land water +0.4 0.2 Ice Sheets +0.6 0.2 Glaciers +0.8 0.4 Sea Level Budget between 1993 and 2010 From IPCC, 2014; Meyssignac & Cazenave, 2012 Sea level rise since 1900 (Meyssignac & Cazenave, 2012) Strong but still uncertain contribution, our goal is to better constrain their mass balance Glaciers: what/where are they? Alaskan Icefields 20 km Vatnajökull (Iceland) Zuo & Oerlemans (1998) Total area = 730 000 km2 Sea Level equivalent (SLE) = 0.5 m Number : >250 000 Mer de Glace Glacier volume changes Berthier et al., Nat Geo, 2010 300 m Retreat and thinning of Columbia Glacier (Alaska) between 1980 (Landsat + USGS DEM) and 2007 (SPOT5) Need for high resolution and accurate DEM to map glacier elevation changes Need for velocity fields to better understand how mountain glaciers are reacting to climate change Pléiades 1A & 1B provide potentially suitable data to accurately estimate topography and velocity Pléiades is also an excellent analog of a future satellite mission (Z-Earth) recommended during the last CNES prospective seminar in La Rochelle Outline of the talk 1. Study sites 2. Radiometric performance over glaciers 3. Accuracy of Pleaides DEMs 4. Velocity mapping from multi-temporal Pléiades images 1. Pléiades data used in this assessment Mostly stereo pairs for DEM generation Also monoscopic acquisitions for velocity mapping (Andes, Alps, Antarctica) Processing with a commercial software (PCI Geomatica 2013) Iceland Univ Iceland, Magnusson, Palson, Gunnlaugsonn Andes IANIGLA, Mendoza Masiokas, Ruiz, Pitte Alpes LGGE Vincent, Rabatel Nepal LGGE Wagnon, Vincent Astrolabe Glacier LGGE Le Meur 1. Pléiades data used in this assessment 2. Radiometric performance Digitization and consequence for the archive SPOT5: 8 bits digitization, DN ranging from 0 to 255. Low gains had to be defined to avoid saturation on snow/ice (high reflectance). Another issue was the limited dynamic of the radiometry in bright homogeneous area of glaciers = narrow histogram of the DN. Consequence : Very few useful data in the archive. Pléiades: 12 bits, DN ranging 0 to 4095. No saturation should occur (gains cannot be tuned). +++ Each image from the catalog is now potentially suitable for glaciology. Our dream? A massive acquisition strategy to populate a vast archive of imagery. (example of ASTER ~20 to 40 DEMs per sites since 2000) Radiometric performance generally excellent. But saturation on the slopes facing the sun and when fresh snow covers the ground Everest (25 November 2012): 3% Mont-Blanc: Between 0.2 and 10% of saturation Iceland: no saturation DN max = 3703 (low sun angles) 2. Radiometric performance t Mont Blanc Mont Blanc Summit Summit Busy on 19 Busy on! 19 August 2012 August 2012 ! 2. Radiometric performance t+32s Mont Blanc Summit Busy on 19 August 2012 ! Radiometric performance Trail to Mont Blanc Summit 2 m accumulation stake Radiometric performance. Saturation Pléiades images of the Mont Blanc on 5 July 2013. Fresh snow and high sun angle lead to 10% saturation (mostly Southeast facing slopes) 3. Vertical accuracy of Pléiades DEM Wagnon et al. The Cryosphere Nov. 2013 First published Pléiades DEM for glaciology (Mera Glacier, Nepal) No accurate ground control points (GCPs). So GCPs derived from coarser resolution SPOT5 DEM and 2.5 m imagery (vertically and horizontally adjusted on kinematic GPS measurements). DGPS measurements performed within +/- 3 days of the acquisitions of the Pléiades stereo-pair: Mean/median dh ( , bias) = -0.93 m Standard deviation ( dh = 1.02 m (N=445) Pléiades images of Mera Glacier and, in blue, location of the GPS measurements acquired to assess the vertical accuracy of the DEM Other similar assessments (with 5-10 accurate GCPs) Andes. Agua Negra. = -0.03 m ( =1.13 m, N>5000) Andes. Tronador. = 0.15 m ( =1.07 m, N=2301) Mont Blanc-2012. = 0.97 m ( =0.67 m, N=491) Mont Blanc-2013. = 0.08 m ( =0.55 m, N=460) Comparison to Lidar data. Case study in Iceland LIDAR data: acquired in August 2011 on Tungnafellsjökull (Iceland). Lidar campaigns to provide a snapshot of the topography all Icelandic ice masses (total area > 10000 km²). Campaigns during 2007-2012. 2 m by 2 m DEMs. Johannesson et al. Annals Glaciol. 2013 Pléiades stereo-pair : 10 October 2013. B/H = 0.45 Processing: 19 GCPs off glacier extracted from a shaded relief image of Lidar DEM Statistics off glaciers (pixels for which dh exceed +/- 10 m are excluded) = -0.3 m ; =1.2 m ; N=3 754 635 Pléiades 10 October 2013 Elevation changes between August 2011 (Lidar) and October 2013 (Pléiades). (on glaciers) = -1.65 m Pléiades DEMs vs. Lidar. Iceland. Necessity of ground control points? Pléiades Shift Shift Mean dh DEM Easting Northing OFF ice (m) (m) (m) Std dev OFF ice (m) Mean dh ON ice 19 GCPs 1.8 -0.7 -0.2 1.1 -1.55 No GCP 2.9 -1.3 2.9 1.1 -1.55 1 GCP 2.0 -0.7 -0.1 1.1 -1.56 Statistics on the horizontal shift and the elevation difference (off and on glaciers) between different Pléiades DEM and the Lidar DEM - Larger horizontal shift without GCP + Mean vertical bias of about 3 m. - One good GCP is sufficient to remove most of the horizontal and vertical biases - After horizontal and vertical adjustment all DEM lead to very similar elevation changes on glaciers Thinning of Mont Blanc glaciers 2003-2012 About 60 m of thinning for the tongue of Mer de Glace and Argentière glaciers Maximun thinning of 110 m in only 9 years! Glacier outlines are from Paul et al., AoG, 2009 Mean ice thinning -10.6 m Mass balance -1 m w.e./a Elevation differences between 21 August 2003 (SPOT5) and 19 August 2012 (Pléiades) Thinning of Mont Blanc glaciers 2003-2012 GPS: -14.4 SAT: -15.9 GPS: -33.4 SAT: -32.6 GPS: -14.9 SAT: -16.2 GPS: -37.0 SAT: -36.5 GPS: -17.0 SAT: -16.0 GPS: -38.6 SAT: -40.0 GPS: -27.8 SAT: -33.3 GPS: -49.2 SAT: -47.0 SPOT5 image of the Mer de Glace (left) and Argentière (right) glaciers and location of the transverse profiles where GPS data are repeated every year by LGGE. GPS: field data; SAT = satellite-derived elevation changes. Mean difference = -0.6 m; stddev = 2.4 m 4. Glacier velocity mapping using repeat Pléiades 2 avril 2012 Animation showing the displacements of surface features on glaciers of the Tronador volcano 4. Glacier velocity mapping using repeat Pléiades 21 avril 2012 Animation showing the displacements of surface features on glaciers of the Tronador volcano 4. Glacier velocity mapping using repeat Pléiades Cartographie des vitesses des glaciers sur le Volcan Tronador (Ruiz et al., in prep) entre le 7 mars et le 2 avril 2012 Pléiades in glaciology: first conclusions Radiometry 12 bits encoding is key for glaciologists Data availability We hope a large archive of Pléiades images will be built We look forward to test SPOT6/SPOT7 stereo data! Vertical accuracy vs DGPS and Lidar data = error of about 1 m (at 1-sigma) with good GCPs Vertical bias can be larger without GCPs, at least with our commercial software. Comparison are underway with soft from IGN and CMLA/CNES. Strong interest of the community to receive DEMs and not raw images (see the SPIRIT project for SPOT5, now more than 80 publications). The scientific community has expressed during the recent CNES prospective the need for a continuous topographic scanning of the Earth surface. Z-Earth is one possible solution Acknowledgments Recette Thématique Utilisateur from CNES Pléiades User group from Astrium Geo-information Service (IANIGLA data) ISIS program from CNES (Icelandic and Astrolabe data) Fundings from TOSCA and from the Pirletta fundation Elevation changes between April 2013 (GPS) – October 2013 (Pléiades) Summer elevation changes vs. Altitude. White dot = field data We are now in position to Map of summer glacier elevation changes Mean dh = -3.1 m (standard deviation = 0.9 m) map the pattern of snow thickness. If the density of the snow layer can be inferred we will be able to measure the seasonal mass balance of glaciers Polar Ice Sheets: Greenland and Antarctica Peninsula Antarctic Ice Sheet : Area = 12 * 106 km2 Mean thickness = 2200 m SLE = ~60 m View of the Arctic region. Greenland ice sheet: SLE = 7 m Ice sheet volume changes Map of surface elevation changes for Antarctica 2002-2010. Flament & Remy, 2012 ERS, ENVISAT, ICESAT altimetry Very weak signal (2 cm/a = .3mm/a SLE) and lack of coverage toward to coast where most of the losses are concentrated Map of surface elevation changes for glaciers of the Larsen B area of the Antarctic Peninsula. Differencing of ASTER and SPOT5 DEMs. Berthier et al., 2012 This small region (about 3000 km² or 1/5000 of the ice sheet) is responsible for 5% of the losses
© Copyright 2024 ExpyDoc