The Removal of Lava Flow Subsidence Signals from deformation interferograms at Nyamuragira Volcano, D.R. Congo Andy Toombs & Geoff Wadge Environmental Systems Science Centre National Centre for Earth Observation University of Reading, UK ESA Cat 1 Projects C1P2880 and AOALO.3667
Nyamuragira Volcano: Geological Context Nyamuragira is a massive, intra-rift, basaltic shield volcano situated on the western branch of the East African Rift System. It is the most active volcano in Africa having erupted 16 times in the last 33 years.
Motivation: Reveal Volcano and Volcano-Tectonic Deformation Signals InSAR data mainly C-Band Coherent surfaces limited to recent lava flows Need to decompose the signals due to: Ground Deformation Atmosphere Noise Lava Flow Subsidence Nyamuragira
Eruptions at Nyamuragira:1991-2006 Caldera
Eruptions at Nyamuragira:1991-2006
Co-eruptive InSAR deformation due to dykes (cm) +3 +35 Towards satellite +24 +10 +28 Caldera N Flank??? NW Flank N Flank and Caldera SW Flank and Caldera 1996 1998 2000 2001 2002 2004 2006 Away from satellite -3-6
Eruptions and Data ALOS FBD ALOS FBS Envisat IS7 Envisat IS2 ERS Asc ERS Des JFMAMJJASONDJFMAMJJASONDJFMAMJJASONDJFMAMJJASONDJFMAMJJASOND 1996 1997 1998 1999 2000 ALOS FBD ALOS FBS Envisat IS7 Envisat IS2 ERS Asc ERS Des JFMAMJJASONDJFMAMJJASONDJFMAMJJASONDJFMAMJJASONDJFMAMJJASOND 2001 2002 2003 2004 2005 ALOS FBD ALOS FBS Envisat IS7 Envisat IS2 ERS Asc ERS Des JFMAMJJASONDJFMAMJJASONDJFMAMJJASONDJFMAMJJASONDJFMAMJJASOND 2006 2007 2008 2009 2010 Red bars = Eruptions of Nyamuragira Purple bar = Eruption of Nyiragongo
Subsidence Rate is a Function of Time Since Emplacement and Flow Thickness Temporal x x Spatial x x
Lava Flow Subsidence Modelling: Coherence 2002 Lava Flow 6 Months after emplacement 18 Months 105 Day coherence images
Lava Flow Subsidence Modelling: Method Define lava flow edges and create mask 2002 Flow
Lava Flow Subsidence Modelling: Method Initial Quality Control of Data Series Stack of masked displacement images Equivalent Stack of coherence images Quality Control 1. Only use pairs >70 days separation to improve signal to noise Quality Control 2. Apply mean coherence test for the lava flow
Lava Flow Subsidence Modelling: Method Take profiles across displacement and coherence data Profiles Downslope Atmospheric phase gradient due to topography Deformation (m) Common Profile Series with a Common Master Increasing Temporal Span Perpendicular profiles Assume flow edges are areas of zero deformation. Normalize profiles. Profile Length (m)
Lava Flow Subsidence Modelling: Method Create corrected surfaces for each interferogram by interpolation Original Data Before interpolation: Quality Control 3. Mask profile data more than 2 STD from the mean. Quality Control 4. Mask profile data where coherence < 0.6. Corrected Surface
Lava Flow Subsidence Modelling: Method Calculate a regression function for each pixel location on the model surfaces Stack of corrected surfaces Subsidence history given by mean deformation values and interval of each interferogram with time. Now able to remove the phase due to lava subsidence for any interferogram
Lava Flow Subsidence Modelling: Results The 1991-3 Lava Flow Field 490 day interferogram Envisat IS2 Mode Dec 2002 Apr 2004 Years 10-12 420 day interferogram Envisat IS7 Mode Dec 2006 Jan 2008 Years 14-16
Virtue out of necessity: better estimates of lava flow thickness and volume If min = 1.5 m and edge = 3 m then Volume = 6.3 x 10 7 m 3 Otherwise default average = 3 m and Volume= 4.1 x 10 7 m 3 2002 Flow Assume flow thickness proportional to subsidence Scale by assuming minimum thickness and edge (field) thickness Extra volume from ponding and compound overflow
Conclusions We have shown that it is possible to map, model and remove the phase attributable to lava flow subsidence. Lava flow subsidence modelling may be a necessary requirement to measure inter-eruption-scale deformation on some volcanoes. We don t yet understand the physical mechanisms of subsidence occurring at a greater rate than thermal contraction alone predicts. We can use subsidence signals to map flow thickness and hence, volume.