Embed Size (px)
Citation preview
!!
!!
!!
!!! !
!!
!!
!!
!!
!!
!!
!!
!
!
!!
!!
!
!
!!
!
!!!!
!!
!!
!!
!!
!!
!!
!!
!!!
!!
!!
!!
!!
@@
$
!!
!!
!!
!!!
!!
!!
!
!!!
!!
!!
!!
!!
!!
!!
!!
!!
! !
! ! !! ! ! ! ! !
25 m
N! ! !! !
Enriquillo Fault:
Riser crest:well locatedapproximate
Activechannel
well locatedapproximateconcealed
T2 T2
T2
T2
T1 T1
T1
T1
T0
6.3 +0.9-1.3 m
View to N
6.6 mT2
T2 T1
T1
B
A
!!
! ! ! ! ! !! ! !
! !
! ! ! !
!
!
!!
N25 m
! ! !Enriquillo Fault: Scarp crest
Active channelfan margin
Offset alluvial fan:
fan apex Offset channel?
well locatedapproximateconcealed
8.6 +2.8-2.5
m
14.1 +3.1-2.4 m
A
B
!!!!!
!!!
!
!
!!
!!
!!
!!!!!
!!
!
!!
!
! ! !!
!!
!!!
!
!!!!!!!!!
!!
!!!!!!!!!!
!!
!
!!
!
!
!!!!
!!!!!
! ! !
! !
Enriquillo Fault:
Riser crest:well locatedapproximate
Activechannel
well locatedapproximateconcealed
T2
T2
T2
T1
T1?
T3
T3
T4
T4
T0
T0
T0
T0
50 m
N
32.2+1.8-3.1 m
C
B
T3
T3
31 m
52 +18-13 m
A
B
C
1 m/pixelDEM (RIT)
0.5 m/pixel DEM (manual)
Jacmel
LéogâneMiragoâne
Petit Goâve
Cayes-Jacmel
Port-au-Prince
St. Michel du Sud
Lamentin
2010fault stepover
0 3015
kilometers
N
73°W
18°30'N
cleardegradedinferred
Enriquillo Fault
72°20'W72°30'W
18°30'N
5 km
Dumay
Chauffard
Fayette
Momance River
Froide River
N
Les Cayes
Aquin
Jacmel
Jérémie
Tiburon
Léogâne
La HotteBaradères
Miragoâne
St. LouisCavaillon
Petit Goâve
Anse-à-Veau
Cayes-Jacmel
Port-au-Prince
Cul-de-Sac plain
St. Michel du Sud
Petit Trou de Nippes
Croix-des-Bouquets
Lamentin2010
17701751
fault stepover
0 3015
kilometers
N
73°W74°W
18°30'N
B
1770 2010Nov 1751A
100 km
IX+VIIIVIIV-VIIII-IVI-II
MM
I Sca
le
Results:
Lidar Viewer - 2.7 billion points
Port-au-PrinceLeogane
Raw Point Cloud Illuminated Point Cloud
Recent Landslide
a) 0.5 m lidar @ sitesb) 1 m RIT lidar (850 km2) c) 30 m ASTERd) 500 m Blue Marble
DEM Layers used here:72°W72°30'W73°W
18°30'N
Port-au-PrinceCarrefourLéogâne Pétionville
Jacmel
PetitGoâve Lidar
Coverage
2010
NORTH AMERICAN PLATE
CARIBBEANPLATE
DOMINICANREPUBLIC
HAITI
CUBA
Port-au-Prince
MUERTOS TRENCH
PUERTO RICOTRENCH
Septentrional Fault
EnriquilloFault
GONAIVE MICROPLATE
To yield a bare-earth surface, cloud must be classified.
Original clas-sification removed 91% of points.
• Examination of this classification in the CAVE revealed many deleted points in the ground plane.
Manual clas-sification removed 30% of points, sup-porting higher reso-lution DEMs.
• Manual classifica-tion was done at sites ~0.12 km2 and took several hours per site using the CAVE and Lidar Viewer.
Project Goals:1. Investigate earthquake behavior of the Enriquillo fault system: • evidence of past surface rupture • explore relationship of earthquakes to fault geometry • investigate implications of the 2010 event on future hazards
2. Test the feasibility of conducting virtual field work during rapid scientific response by using new virtual-reality based methods to analyze massive lidar data.
5.1 Map regional fault geometry.
Virtual-Reality Approach:
Motivation:
immersive: user is surrounded by a virtual scale model of the data.interactive: user immediately sees results of data operations.real time: maintain interactive stereoscopic rendering (50 frames/sec).
Use immersive, interactive, real-time data visual-ization to exploit the human capacity for visually iden-tifying meaningful patterns in noisy data.
Our system incorporates both hardware and software that together create a virtual landscape in which users can make many of the same observations that normally require fieldwork.
We optimize data handling so datasets many times larger than main memory can be visualized and interactively manipulated. Specifically: Hierarchical data structures View-dependent, multi-resolution, out-of-core rendering
Software runs on wide range of hardware: standard desktop computers to fully immersive visualization environments such as the CAVE
Methods: Crusta:Lidar Viewer: CAVE:Characteristics:• Virtual globe based on a 30-sided polyhedron = no polar distortion. • Renders arbitrarily large (e.g., whole earth) sub-meter resolution DEM and
image data. • Independent of performance of computer’s graphics subsystem.
Characteristics:• Four-sided (3 walls & floor), 800 ft3 MechDyne
CAVE.• Stereoscopic images projected on all 4 surfaces.• Wireless Intersense IS-900 tracking system syn-
chronizes display to user’s view and tracks input device (six-button handheld wand).
• User perceives seamless 3D environment.
Characteristics:• Enables real-time visual analysis of multi-billion-point
lidar point clouds. • No sub-sampling or reduction in data resolution.• Uses octree data structure to enable multi-scale data
representation and efficient, localized point data re-trieval.
For remote mapping on lidar DEMs (Bernardin et al., 2010) Generates a 3D virtual scale model that users interactively manipulate.For analysis of lidar point
clouds (Kreylos et al., 2008)Functions:• Interactive illumination of the point cloud using multiple light
sources from arbitrary directions (e.g., hillshaded clouds).• Interactive point selection and extraction. • Quantitative point cloud comparison (e.g., full vs. bare-earth).• Fitting geometric primitives (line, plane, sphere, cylinder) to se-
lected points. • Visualization of point distances from a plane. • Extraction of profile curves.
Functions:• Map directly on the virtual landscape using standard geologic anno-
tation and standard GIS formats (shapefiles, kml, etc.).• Hardware-accelerated manipulation of the texture color palette.• Interactively adjust illumination and vertical exaggeration.
Operations used here:• Compare full and bare-earth clouds.• Select points for manual vegetation removal.• Find & evaluate landform offsets.• Measure landform offsets.• Map on virtual terrain.
Our system is an analytical instrument that allows users to discover meaningful patterns in massive, complex datasets, and then measure and record
these observations.
The Model & Implications:
Acknowledgements:
Fayette: Chauffard (Jameau): Dumay:
Procedure:• Map trace & find sites.• Manually remove vegeta-
tion points at sites. • In Crusta: map, project
landforms, and measure offset and uncertainty.
• Export mapping to Arc.• Compare with field obser-
vations (if available).
• Jan. 12, 2010 Mw 7.0 Haiti earthquake was deadliest of its magnitude in recorded history (Bilham, 2010).
• Response effort produced 1st large-footprint (850 km2) lidar dataset collected within days of an earthquake.
• Lidar collection and processing coordinated by Roch-ester Institute of Technology with help/funding from USGS and World Bank. We commend these efforts, particularly the rapid data release.
• Data comprise >2.7 billion lidar points (66.8 GB).
• Enriquillo fault is active, left-slip plate-boundary (Mann et al., 1995, 2002).
• Cluster of 3 earthquakes in 1751 and 1770 (Scherer, 1912).
• GPS data had highlighted earthquake risk (Manaker et al., 2008).
• Our mapping indicates a fault stepover near 2010 rupture (white box).• We see evidence of recent rupture to east of, but not within, the step.
Our mapping revealed clear (red) and degraded (black) geomorphic evidence of recent surface rupture. Inferred areas (white) result from erosional modification/destruction of the fault trace.
5.3 Conduct virtual field studies to find, document, and measure landforms offset during past surface ruptures.
Original Classification
Manual (Lidar Viewer)
Classified point clouds: vegetation vs. ground 91% removed
30% removed
m-2.4+3.1
14.1
m-2.5+2.8
8.6 m-1.3+0.9
6.3m-3.1+1.8
32.2m-13
+1852
2
3
1
5
4
6
7
5.2 Manual vegetation removal.
Shaking effects from our analy-sis of damage reports from 1751 & 1770 (Scherer, 1912) compared to 2010 values.
Our inferred fault rupture pattern (yellow).
DatasetFull cloud (original)Example site (original)Example site (manual)
Total Points2,700,563,734
204,801204,801
Removed (#)2,369,509,911
185,81360,452
(%)889130
Results from point set comparison analysis in Lidar Viewer:
Plate-boundary context.
Virtual field observations from Crusta. A) Site map of 8-10 m wide alluvial fan showing 8.6 m fault offset. Channel to west may be offset 14 m. B) Crusta images with red circles indicating de-flected contour lines defining the offset fan mapped in A.
A) Virtual field observations from Crusta showing fluvial terrace riser offset ~32.2 m. B) Indepen-dent field observations confirm this offset, with mapping (not shown) replicating our virtual field results except for details at the base of the T3-T1 riser. C) Crusta image showing 52 m deflected channel. We tentatively interpret this as a fault offset because the uphill-facing fault scarp may have deflected the stream.
A) Virtual field observations from Crusta showing 6.3 m offset of faulted fluvial terrace riser. B) In-dependent field observations confirm this offset, with field-based mapping (not shown) replicating our virtual field results. We found this site during remote analysis and shared the location with the USGS field party, leading to the field observa-tions.
1 m contours
Airborne lidar imaging to support emergency response and damage assess-ment in urban areas was coordinated by the Rochester Institute of Technology (RIT)
with funding from the World Bank (WB). The lidar data were collected starting ~10 days of the earthquake and released as a bare-earth digital elevation model (DEM) and clas-sified point clouds within a week of collection. Such rapid and open data release serves as a model for future rapid response and proved essential for the analysis re-ported here. In detail., the lidar data were collected by Kucera International in collabo-ration with ImageCat. The U.S. Geological Survey coordinated expansion of the sur-veys to encompass the Enriquillo fault south of Port-au-Prince from L'acul to Lago En-riquillo. Classification of the data to produce a bare-earth DEM was conducted by the Information Products Laboratory for Emergency Response (IPLER; http://lias.cis.rit.edu/projects/ipler). We infer that this classification used a very low point-to-point threshold angle to identify the bare earth, based on the frequency of points misidentified as vegetation. While appropriate in the low-relief urban regions
primarily targeted by the survey, this scheme performed poorly in the steep terrain along the Enriquillo fault.
Bernardin et al., 2010, Comp. Geosci., in press.Bilham, 2010, Nature, v. 463, p. 878.Kreylos et al., 2008, Adv. Vis. Comp. ISVC 2008,
Lec Notes Comp. Sci., p. 846.Hanks and Bakun, 2002, Bull. Seis. Soc. Am., v.
92, p. 1841.Hanks and Kanamori, 1979, J. Geoph. Res., v. 84,
p. 2348.Manaker et al., 2008, Geoph. J. Int., v. 174, p.
889.Mann et al., 1995, Tectonophysics, v. 246, p. 1.Mann et al., 2002, Tectonics, v. 21, p. 1057.Prentice et al., 2010, Nat. Geosci., v. 3, p. 789.Scherer, 1912, Bull. Seis. Soc. Am., v. 2, p.
Model for the earthquake behavior of the central Enriquillo fault
Implications for seismic hazard
Our system transforms lidar analysis during rapid response• This is the first reported use of data visualization in an immersive virtual-reality environment to analyze massive multi-resolution terrain data.• The advanced methods of virtual-reality based data visualization we report here enable virtual field stud-ies and real-time interactive analysis of massive lidar datasets.
• Fault is geometrically segmented by a stepover between L’acul and Fayette.• We infer the fault ruptured east of the step in 1751 and to the west in 1770. • The 2010 earthquake was a relatively small event within the step between the earlier earthquakes.
• Virtual fieldwork yields evidence of protracted surface rupture between Fayette and Dumay, 33 km apart.• An earthquake similar to, or larger than the 2010 event is possible along this portion of the fault:
- We estimate past surface rupture was 40 to 60 km along strike, based on fault morphology locations where we see 6-8 m offsets. - Theoretical relationships based on rupture area alone (Hanks and Bakun, 2002) suggest magnitudes of Mw 6.8 to 7.0. - Magnitudes Mw 6.9 to 7.2 are implied by Hanks and Kanamori, 1979, using 40-60 km rupture with mean slip of 2-3 m (Prentice et al., 2010).
References:
1 Dept. of Geology, Univ. of California, Davis, CA 956162 IDAV, Dept. of Computer Science, Univ. of California, Davis, CA 956163 United States Geological Survey, Denver, CO, 802254 United States Geological Survey, Sacramento, CA, 95819
Earthquake behavior of the Enriquillo fault zone, Haiti revealed by interactive terrain visualizationU13A-0015
Eric Cowgill1, Tony S. Bernardin2, Michael E. Oskin1, Christopher Bowles1, M. Burak Yıkılmaz1, Oliver Kreylos2, Austin Elliott1, Scott Bishop2, Ryan D. Gold3, Alexander Morelan1, Gerald W. Bawden3, Bernd Hamann2, Louise H. Kellogg1
