Processing satellite imagery for mapping physical exposure globally
1. Processing satellite imagery for mapping
physical exposure globally
Ehrlich D., Halkia S., Kemper T., Pesaresi M., and Soille P.
Session: Global exposure monitoring for multi-hazard risk
assessments
4TH INTERNATIONAL DISASTER AND RISK
CONFERENCE - IDRC DAVOS 2012
2. Why satellite imagery?
• Imagery for quantifying physical exposure
• Abundance of imagery
• Exposure maps from imagery
• Process large volume of data
• Global – cities and rural areas
• New processing systems in place
4. Why is Google Earth not sufficient?
• Images are not Data
enough Data
• Buildings, roads, trees …
• We need numbers (digital
maps) Information
• How many buildings?
• What is the extent of Information
settlements
• How much is at risk?
5. Where, how much, how many?
Techniques:
1. Manual encoding (above)
2. Machine assisted (train a computer algorithm - automatic) next slides
8. Human Settlement derived from
SPOT-5
The image is processed to
generate a map that
contains information on
human settlements, i.e.
Alger density of
buildings, number of
buildings
This is just DATA
Digital Map
(yellow)
This is INFORMATION
9. Human settlement derived from
Landsat
Settlement maps for
1.London
2.Delhi
3.Los Angeles
4.Paris
5.Roma
6.San Francisco
7.Jakarta
8.Madrid
9.Milan
Each image
shows 36 x 36 km
All cities of the
world could be
mapped
11. New concept: Human settlement
analysis and monitoring system
• Take all imagery necessary
• Use standardized algorithms that work across the
globe on a number of image types
• Put in place an infrastructure that can process
imagery covering the entire Earth’s land masses
12. Information flow for the global
VHR: Ikonos,
QuickBird, human settlement system analysis
World View…
and monitoring system
High Res
SPOT, CBERS
Human Human
Settlement Physical
Medium Res Settlements
Indices exposure
Landsat (Global)
Coarse Res
i.e. Modis
Vulnerability
Information
MODIS-Urban
Landscan
13. VHR complexity: data size
Number of pixels needed to cover 1 sq km
4,500,000
WorldView-1 P
4,000,000
3,500,000
3,000,000
QuickB. P
2,500,000
2,000,000
1,500,000
IKONOS P
1,000,000
Landsat MSS Landsat TM Spot4 XS Spot4 P
500,000
Spot5 P
0
50 45 40 35 30 25 20 15 10 5 0
IRS-1 IKONOS MS QuickB MS
Sensor spatial resolution (m)
23. Global – All settlements
Sparse settlements as found in rural areas are often not
accounted for in exposure mapping
24.
25.
26.
27.
28.
29.
30. New VHR GHSL Model
Panchromatic image of Sana’a Yemen (8956×16384, 8-bit elements).
Tree computation time (both): 50sec. Layer computation time: 17sec. No Free Parameter – blind computational
allowed
31.
32.
33. Earth Observation technology
(satellite images) for exposure
• Very rich source of information on the Earth
surface
• More sensors will be launched and more imagery
will be available in the future
• The detail of the imagery is adequate to map
physical exposure globally
• Extracting information is costly but new algorithm
can facilitate the process
• Computing power is no longer a limitation
34. JRC will run two services for the
community (from web portal)
1. Information on demand
• Send area of interest and receive information
A. If coarser imagery is available
A. Density of built up
B. If Very High Resolution imagery is available
A. Density of built up
B. Number of buildings
C. Average size of buildings
2. Processing on demand
• Send imagery, and receive the information
35. Thank you for your attention
ISFEREA Action
European Commission • Joint Research Centre
IPSC/Global Security & Crisis Management Unit
Tel. +39 0332 785648
Email: ghslsys@jrc.ec.europa.eu