matsuoka_igarss2011.ppt

18
Coseismic Displacement Measurement of the 2010 El Mayor, Mexico Earthquake by Subpixel Correlation from Optical Satellite Images Masashi Matsuoka and Shinsuke Kodama AIST, Japan

Upload: grssieee

Post on 27-Jun-2015

276 views

Category:

Technology


0 download

TRANSCRIPT

Page 1: matsuoka_igarss2011.ppt

Coseismic Displacement Measurement of the 2010 El Mayor, Mexico

Earthquake by Subpixel Correlation from Optical Satellite Images

Masashi Matsuoka and Shinsuke KodamaAIST, Japan

Page 2: matsuoka_igarss2011.ppt

Objective

• In order to grasp the quantitative and spatial information of the earth surface deformation due to earthquakes, both interferometric analysis from phase information and disparity analysis by subpixel correlation from images using SAR data are well-known.

• The latter approach can be applicable to recent high-resolution optical images.

• This presentation performs to apply the subpixel-based image processing using pre- and post-earthquake orthorectified images to map the displacement distribution around the fault rupture region of the 2010 El Mayor (Baja) earthquake.

Page 3: matsuoka_igarss2011.ppt

InSAR and Pixel-offset- InSAR

merit:• centimeter accuracy (microwave length)demerit:• difficult to measure fault displacement directory because of

large amplitude and low coherence near surface rupture zone• applicable to SAR

- Disparity (Pixel offset) analysismerit:• robust measurement • applicable not only SAR but optics as welldemerit:• meter accuracy (approx. one to ten in horizontal (range)

resolution)

Page 4: matsuoka_igarss2011.ppt

The 2005 Pakistan Earthquake

2000/11/14 2005/10/27

R: Band3NG: Band2B: Band1

ASTER images (resolution: 15 m) taken pre- and post-event images

The Mw 7.6 earthquake struck Northern Pakistan and Kashmir on October 8, 2005.

Avouac et al. (2006)

R: Band3NG: Band2B: Band1

Page 5: matsuoka_igarss2011.ppt

Displacements from Pixel-offset of ASTER Images

East-west ground displacement(white to the east, black to the west)

Avouac et al. (2006)

South-north ground displacement(white to the south, black to the north)

Page 6: matsuoka_igarss2011.ppt

Disparity (Pixel-offset) Analysis

Page 7: matsuoka_igarss2011.ppt

Calculation Flow

1. Tie points:are obtained from the optimal pixel pair selection, by searching sub-pixel by sub-pixel, at the position that yields the highest cross-correlation of the area of the local window between the slave (post-event) and the master (pre-event) images.

> Interval: 10-pixel> Local window size: 31 31 pixels> Threshold of correlation coefficient: 0.7

2. Slave image coregistration:is carried out by the second degrees polynomial regression using the points, whose correlation coefficient is larger than 0.95, in the above selected tie points.

3. Disparity:can be calculated from the residual in the locations of original and coregistered pixels for x (east-west) and y (south-north) directions of slave image.

Page 8: matsuoka_igarss2011.ppt

Baja California, Mexico EarthquakeDate: April 4, 2010Magnitude: 7.2

Page 9: matsuoka_igarss2011.ppt

Pre- and Post-earthquake PRISM Images

2009/09/14 2010/05/02

10 km 10 km

Path: 542 (west)

Page 10: matsuoka_igarss2011.ppt

ALOS/PRISM Capabilities

(source: JAXA)

OB1

OB2

DEM generation

Ortho-rectification

Pixel resolution: 2.5 m

Subpixel accuracy:approx. quarter meter

Page 11: matsuoka_igarss2011.ppt

Pre- and Post-earthquake ortho-rectified PRISM Mosaic Images

before after

Page 12: matsuoka_igarss2011.ppt

Displacements from Pixel-offset of PRISM Images

east-west component south-north component

10 km 10 km

− + −

+

The correlation images were obtained with a sliding 31×31 pixels correlation window and 10-pixel step. Spatial resolution of the correlation images is approximately 75 m. Decorrelation areas where the cross-correlation coefficient is lower than 0.7 are discarded and appear in white. right-lateral strike slip fault

Page 13: matsuoka_igarss2011.ppt

Displacements from Pixel-offset of PRISM Images

10 km 10 km

− + −

+

east-west component south-north component

The correlation images were obtained with a sliding 31×31 pixels correlation window and 10-pixel step. Spatial resolution of the correlation images is approximately 75 m. Decorrelation areas where the cross-correlation coefficient is lower than 0.7 are discarded and appear in white.

Page 14: matsuoka_igarss2011.ppt

East-west Component from PRISM Image Pixel-offset

Six profiles, A-A’ to F-F’, run across the fault trace are selected.

Displacements are positive toward the east.

Slip vectors of the east-west component of ground displacement obtained by six profiles.

Page 15: matsuoka_igarss2011.ppt

South-north Component from PRISM Image Pixel-offset

Six profiles, A-A’ to F-F’, run across the fault trace are selected.

Displacements are positive toward the south.

Slip vectors of the south-north component of ground displacement obtained by six profiles.

Page 16: matsuoka_igarss2011.ppt

Stacked Profiles

East-west component South-north component

It can be clear to recognize the discontinuity caused by the fault rupture and the maximum displacements of approximately 2 m and 1.5 m are measured in the east-west and south-north directions, respectively.

Page 17: matsuoka_igarss2011.ppt

Comparison with Field Survey

(Geotechnical Earthquake Engineering Reconnaissance, GEER Association Report No. GEER-023)

The extracted displacements from PRISM image correlation are comparable to field survey results.

Page 18: matsuoka_igarss2011.ppt

Conclusions• This paper performed to measure the ground deformation of

the fault rupture due to the 2010 El Mayor, Mexico earthquake by the sub-pixel correlation technique to pre- and post-event satellite orthorectified images. The ALOS/PRISM images was used for this examination.

• The maximum displacement of right-lateral strike slip interpreted from the image analysis was approximately 2 m and 1.5 m in the east-west and the south-north components, respectively.

AcknowledgementThe PRISM data, which is owned by JAXA, was provided by Jishin Working Group under the project of MEXT. The data was processed by the GEO Grid, AIST.