Slide 1

Comparison of L and P band radar time series for the monitoring of Sahelian area P.-L. Frison, G. Mercier, E. Mougin, P. Hiernaux Slide 2 Context: Better understanding Sahelian surface processes and their interaction with monsoon variability Improve our understanding and documentation of long term trend in vegetation in response to climate change radar data: 2 key parameters: soil moisture and vegetation Goal: Comparison of L band PALSAR and C band ASAR data for Sahelian surface monitoring. Relation between radar vs surface parameter temporal evolution Slide 3 Outline: Study site PALSAR and ASAR data change detection method Results and discussion Slide 4 The Sahel Semi-arid area herbaceous layer (0-50 %) (annual grasses) Dry season (Nov. Apr.) bare soil Rainy season (May Oct.) shrub (0-20 %) trees (1-5 %) + Slide 5 Region of study: the Gourma - Mali Seno Slide 6 PALSAR acquisitions: L band (Jan. 2007 - Apr. 2009) DATASET Mode#PolarizationResolutionIncidence angle Pass Fine Beam7HH15 m40Ascending Fine Beam5HH / HV15 m40Ascending Wide Swath6HH100 m30Descending ASAR acquisitions: C band (Jul. - Dec. 2005) Mode#PolarizationResolutionIncidence angle Pass Wide Swath41HH150 m18-35Ascending + Descending Slide 7 Gourma Region (MALI) ASAR Wide Swath - HH 20 th Dec 2005 Slide 8 GOURMA Region (MALI) PALSARWIDE BEAM- HH 1st Jan 2008 Slide 9 Gourma Region (MALI)ASAR C-band PALSAR L-band C-band (ASAR): Shallow sand and silt soils L -band (PALSAR): Better discrimination of geological features Remnant of alluvial systems and lacustrine depressions Slide 10 Gourma Region (MALI)ASAR C-band PALSAR L-band C-band (ASAR): Shallow sand and silt soils L -band (PALSAR): Better discrimination of geological features Remnant of alluvial systems and lacustrine depressions Slide 11 PALSAR Fine Beam HH polarization Temporal color composite image GOURMA - MALI 17 Jan. 2007 20 Oct. 2007 22 Jan.2009 Water ponds Hombori mounts Low-land (accacia forest) Slide 12 Change detection method Constraints : Large dynamic range (high differences over bright patterns) Even after multi-looking, presence of noise (speckle) absolute or relative differences, ratios, rms,.. not significant Time series color composite imageRelative differences Slide 13 Temporally stable regions gray Change detection colored areas Case of 3 channels Change detection method Constraints : Large dynamic range (high differences over bright patterns) Even after multi-looking, presence of noise (speckle) absolute or relative differences, ratio, rms,.. not significant Slide 14 Temporally stable areas gray areas (no saturation) Change detection colored areas (saturation) RGB space R B G 0 HSV space Case of 3 channels Value Saturation Hue Change detection method Constraints : Large dynamic range (high differences over bright patterns) Even after multi-looking, presence of noise (speckle) absolute or relative differences, ratio, rms,.. not significant Slide 15 17 Jan. 2007 20 Oct. 2007 22 Jan.2009 RGB Space Slide 16 Value Hue Saturation HSV Space Areas that have changed Slide 17 Color composite image Saturation image 17 Jan. 2007 20 Oct. 2007 22 Jan.2009 Change detection for a 3-date color composite image PALSAR Fine Beam HH polarization Slide 18 Change detection method Case of N channels (N>3): P iterations: 1) random draw of 3 among the N available channels 2) Compute the saturation channel from HSV space Average of the P saturation channels Example: 12 Finebeam acquisitions at HH pol. N=12 12! / (9! * 3!) = 220 possible random draws P =50 (arbitrary) Slide 19 Color composite image Saturation image 17 Jan. 2007 20 Oct. 2007 22 Jan.2009 Change detection for a 3-date color composite image PALSAR Fine Beam HH polarization Slide 20 Color composite image Saturation image Change detection for 12 FineBeam acquisitions (HH polarization) 17 Jan. 2007 20 Oct. 2007 22 Jan.2009 PALSAR Fine Beam HH polarization Jan. 2007 Apr. 2009 Slide 21 Temporal changes detected over 12 Fine Beam acquisitions PALSAR data HH polarisation Jan. 2007 Mar. 2009 Water ponds Fields (millet) dep. Orientation! Significant penetration depth over sandy soils Slide 22 Changes in dry season Changes in rainy season Temporal changes detected over 12 Fine Beam acquisitions Water ponds permanent PALSAR data HH polarisation Jan. 2007 Mar. 2009 Fields (millet) dep. Orientation! Slide 23 ALOS/PALSAR FBD 6 th June 2008 HH HV Sandy soils Shallow soils INFLUENCE OF POLARISATION Slide 24 Change detection between HH and HV PALSAR DUAL POLARIZATION Shallow soils + Water ponds Slide 25 ALOS/PALSAR WB1: HH Polarization Water ponds discrimination 2008 dry season: 1 Jan 16 Feb 2 Apr 3-date color composite image Slide 26 Change detection (dry season) ALOS/PALSAR WB1: HH Polarization Water ponds discrimination Main water resource Hydrological indicator surface runoff areas increase since begining of drought period (50s) Slide 27 ASAR Wide Swath HH polarisation 41 acquisitions 22 acquisitions in ascending pass 5 acquisitions same incidence angle (35) Jul. Dec. 2005 Slide 28 ASAR Wide Swath - HH polarisation 2 nd Sept. 2005 change detection (5 dates) low penetration over sandy soil upper surface changes sandy soils Slide 29 ASAR Wide Swath - HH polarisation 2 nd Sept. 2005 change detection (5 dates) low penetration over sandy soil upper surface changes sandy soils, water ponds Slide 30 change detection (5 dates) PALSAR (L-band) low penetration over sandy soil upper surface changes sandy soils, water ponds High penetration dpeth over sandy soils water ponds millet fields Comparison P band / L band temporal change detection Slide 31 Conclusion RGB H S V simple but performant for change detection ! Temporal change detection: penetration depth illustration histogram must be the same across the hole image! C band (low penetration over sandy soils) mix of upper surface changes sandy soils ( soil moisture + vegetation) water ponds more difficult te discriminate special features L band: some variation over sandy soils (soil moisture?) Cross over with SMOS mission L band (high penetration over sandy soils) Water ponds Millet fields Slide 32 Thank you for your attention!