REMOTE-SENSING ASSESSMENTOF WIND DAMAGE

J. Arn Womble, Ph.D., P.E. Wind Engineering Consultant, LNSS

Instructor, Texas Tech University

Beverley J. Adams, Ph.D. ImageCat, Inc.

Kishor C. Mehta, Ph.D., P.E. Texas Tech University

DigitalGlobe, Inc. DigitalGlobe, Inc.

Wind Science and Engineering

ImageCat, Inc.

Windstorm Damage Descriptors

Missile Impacts

(Debris)

Wall Failures

Door, Window Failures

Roof Structure

Roof Deck

Roof Covering

Oklahoma City Tornado (TTU-WISE Aerial Photo, 1999)

Broad range of damage visible from above

Residential Damage States HAZUS Hurricane (FEMA)

• Rapid Assessments (data at many levels)

• Preservation of Damage Conditions

• Access

• Debris

• Areawide Screening for Detailed Surveys

• Uniform (Unbiased) Assessments

Improved Damage Assessmentswith Remote Sensing

Data Availability

QuickBird Images – Southeast U.S.

(July 2005)

(www.DigitalGlobe.com)

(May 2007)

(Biloxi, Mississippi – Hurricane Katrina)

NOAA Aerial Images – Katrina

Credit: Google Earth, NOAA

Credit: Google Earth, NOAA

• Aug. 29, 2005 – Hurricane Katrina

• Aug. 30, 2005 – First image collections

• Aug 31, 2005 – Raw images available online

(not georeferenced)

• Georeferenced via GoogleEarth weeks later

• 37 cm resolution

• Free!

Pictometry (Oblique Aerials)

© 2005 Pictometry International

Satellite vs Aerial Imaging

... Relative merits are changing rapidly

... Depends on situation (case-by-case thus far)

... Ideally, be prepared to use either (or combination)

Pre-Storm

Imagery

(archives)

Access

(weather, orbits,

flight restrictions)

Collection Speed

Coverage

Area

Spatial Resolution

Flight plans

Georeferencing

CostStereo Imaging

Use of Remote-Sensing Data

Linked by

RS Damage Scale

& Field Studies

QualitativeVisual, Semi-Automated

QuantitativeAutomated (Change Detection)

Damage metrics

21 Sensors

Resolutions:

5 cm to 5 km

Qualitative &

Quantitative

Usage

MCEER

Exploratory Study (QuickBird Satellite Imagery)

Changes in Pre-/Post-storm imageryField Observations

Hurricane CharleyAugust 13, 2004

Hurricane IvanSeptember 16, 2004

Credit: NOAA

Qualitative Review – Building Damage

• We can visually detect damage – HOW?

• What changes can the computer detect?

Remote-Sensing Damage “Signatures”

• Degradation, appearance of edges (perimeter, interior)

• Removal or appearance of materials (decking, roof cavities, framing members)

• Texture (orderliness or randomness)

• Morphology (shapes)

• Presence of debris

Field Data Collection

GPS Tracks

• >10,000 Buildings

• Hurricane CharleyPunta Gorda, FLPort Charlotte, FL

• Hurricane IvanPensacola, FL

Credit: ImageCat, Inc. and DigitalGlobe (www.digitalglobe.com)

QuickBird Images (Hurricane Charley)QuickBird Natural Color (03-23-04)

DigitalGlobe, Inc.

QuickBird Natural Color (08-14-04)DigitalGlobe, Inc.

BEFORE

AFTER

Credit: DigitalGlobe (www.digitalglobe.com)

Building Types

(Size and Complexity)

Large Industrial Buildings

NOAADigitalGlobe (62 cm)

DigitalGlobe (80 cm)DigitalGlobe (62 cm)

DigitalGlobe (62 cm) NOAA

Residential Buildings

DigitalGlobe (61 cm)

DigitalGlobe (61 cm)

DigitalGlobe (61 cm)

DigitalGlobe (61 cm)

DigitalGlobe (75 cm)

DigitalGlobe (75 cm)

Residential – Spatial Resolution

QuickBird Satellite (61 cm)

QuickBird Satellite (75 cm)

Digital Aerials(DMK Associates)(5 cm)

BEFORE

AFTER

AFTER

Credit: DigitalGlobe (www.digitalglobe.com)

Credit: DigitalGlobe (www.digitalglobe.com)

Credit: DMK Associates

Mobile Homes

DigitalGlobe (61 cm)

DigitalGlobe (61 cm)

DigitalGlobe (61 cm)

DigitalGlobe (75 cm)

Credit: DigitalGlobe (www.digitalglobe.com)

Credit: DigitalGlobe (www.digitalglobe.com)

Remote Sensing Damage Scale

Quantitative Damage Description

• Need automation for• Speed• Consistency

• What changes can the computer detect?

• Must express changes numerically...effectively

• Correlation with actual damage assessments

DigitalGlobe, Inc.

Pre-Storm (3-23-04)DigitalGlobe, Inc.

Post-Storm (8-14-04)

Object-Based Damage Metrics

Use of individual roof facets (slopes):

• Illumination differences• Aerodynamic “discontinuities”• Object extraction

Object-level analysis(not pixel level)

Allows use of different• pixel sizes • number of pixels/objects• view angles• spatial resolutions• platforms• alignment (location)

Windstorm Damage Profiles (Residential)Idealized Damage Profile

• Correlates Damage Metric & Damage States

• Unique Correspondence for Each Damage State

• Basis for Computer Algorithms (Automation)

Damage States

Damage M

etric

No Damage

Maximum Damage

Da

ma

ge M

etr

ic

STANDARD DEVIATION DIFFERENCE (after - before) [Linear Scale]

Green Band

-200

-150

-100

-50

0

50

100

150

200

250

300

1

A B C D

Damage State

Blue Band

-200

-150

-100

-50

0

50

100

150

200

250

300

1 1 2 2 3 3 4 4

A B C D

Damage State

Red Band

-200

-150

-100

-50

0

50

100

150

200

250

300

1 2 3 4

A B C D

Damage State

NIR Band

-200

-150

-100

-50

0

50

100

150

200

250

300

1 2 3 4

A B C D

Damage State

A B C D

UNIFORMITY RATIO (after / before) [Linear Scale]

Dam

ag

e M

etr

ic

0.0

0.5

1.0

1.5

2.0

2.5

1 2 3 4

A B C D

Damage State

0.0

0.5

1.0

1.5

2.0

2.5

1 2 3 4

A B C D

Damage State

0.0

0.5

1.0

1.5

2.0

2.5

1 2 3 4

A B C D

Damage State

0.0

0.5

1.0

1.5

2.0

2.5

1 2 3 4

A B C D

Damage State

A B C D

Windstorm Damage Profiles (Residential)

Metal WarehousesD

am

ag

e M

etr

ic

STANDARD DEVIATION DIFFERENCE (after - before) [Linear Scale]

-100

0

100

200

300

400

500

1 2 3 4

A B C D

Damage State

-100

-50

0

50

100

150

200

250

300

350

1 2 3 4

A B C D

Damage State

-50

0

50

100

150

200

250

300

350

400

1 2 3 4

A B C D

Damage State

-100

-50

0

50

100

150

200

250

300

350

1 2 3 4

A B C D

Damage State

Dam

ag

e M

etr

ic

UNIFORMITY RATIO (after / before) [Linear Scale]

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

1 2 3 4

A B C D

Damage State

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

1 2 3 4

A B C D

Damage State

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

1 2 3 4

A B C D

Damage State

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

1 2 3 4

A B C D

Damage State

A-11; B-21; C-17; D-25

Future Steps & Needs

Need for Rapid Imaging

Aug. 14 Aug. 19Mar. 12

Debris....Cleanup....Repairs....Roof Covers

DigitalGlobe

Edges are Very Important (High Resolution)

(Tom Smith, AIA, 1992)

DigitalGlobe

Object Extraction

DigitalGlobe, Inc.

Pre-Storm (3-23-04)DigitalGlobe, Inc.

Post-Storm (8-14-04)

Material Boundaries

NOAA NOAA

Edge Detection (Laplacian Filter)

DigitalGlobe, Inc.

DigitalGlobe, Inc.

Credit: DigitalGlobe (www.digitalglobe.com)

Spectral Analysis – Materials (Hyperspectral Beyond Color)

False Indicators of Damage

Spectral Signatures ....

• Water ponding on roof

• Overhanging Vegetation (stripped at low windspeeds)

• Solar collectors

Debris Mapping/Characterization

Material Separation

Need for 3-D Data (Heights)

Photo Credit: NOAA Photo Credit: ImageCat, TTU-WISE

SAR & LIDAR

can help!

LIDAR & SAR

NASA EAARL and USGS

• All Weather/Light Conditions

• Elevations (3-D Data)

NASA EAARL and USGS

Applications to Hurricane Katrina Damage

(2005 – present)

Applications to Hurricane Katrina - 2005

“Slab” Cases

Neighborhood Damage Patterns

NOAA 37-cm digital aerial images acquired Aug. 30, 2005

N

Subject

Residence

Dense Surge Debris

Partial Loss of

Roof Covering

Displaced

buildings or

roof

assemblies

N

Subject

Residence

Pre-storm baseline imagery

VIEWS – Google Earth (Katrina 2005)

Credit: ImageCat, MCEER, and Google Earth

Subject

Blue Tarps – Recovery Imagery

Indicator of wind damage ...

... Damage obscured, but limited.

Conclusions

• Qualitative Study• Ability to assess damage depends on

• Building size• Geometric complexity• Sensor resolution

• Building types – different damage signatures• Remote Sensing Damage Scale (Residential Buildings)

• Roadmap to automated damage assessment identified• Understanding of how humans perceive damage• Understanding of how computer sees change (“damage”)• Characterize windstorm damage in terms of remote-sensing signatures• Correlation of computer damage metrics with actual damage states

• Quantitative Study (Correlations: Damage Metrics & Actual Damage States)• General damage trends are apparent• Current satellite resolutions (61cm) not sufficient to make definitive correlations• Need enhanced spatial (and spectral) resolutions• Need to explore additional damage metrics, combinations

(QuickBird Imaging Satellite, www.DigitalGlobe.com )

REMOTE-SENSING ASSESSMENTOF WIND DAMAGE

J. Arn Womble, Ph.D., P.E. Wind Engineering Consultant, LNSS

Instructor, Texas Tech University

Beverley J. Adams, Ph.D. ImageCat, Inc.

Kishor C. Mehta, Ph.D., P.E. Texas Tech University

DigitalGlobe, Inc. DigitalGlobe, Inc.

Wind Science and Engineering

ImageCat, Inc.

Hyperspectral Examples

NASA EO-1

• Hyperion sensor

• Acquisition Date: May 1, 2002

• Spatial resolution: 30-meter

• Spectral bands: 242 (198 calibrated)

• Spectral coverage: 356-2577nm (427-2395nm for calibrated bands)

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

400 600 800 1000 1200 1400 1600 1800 2000 2200 2400

Point 3a

Point 3b

Point 3c