Extraction of Exposure Information from Earth observation

The discussion will focus on extracting “physical exposure” or “physical exposed assets” as used in the disaster risk equation (Dilley et al. 2005) from Earth Observation data.

Welcome to the EO for exposure group (EO and exposure).

The group will discuss (1) EO dataset, (2) informaiton extracton techniques to map(3) physical exposure. With physical exposure we refer to the man made structures (physical assets) that make up settlements, towns and cities as seen from Earth Observation (EO).

SUMMARY OF DISCUSSIN TO DATE
1. What EO data?
What Earth Observation data has been used or could be used to extraction exposure information?
Imagery used to quantify exposure at local scale:
a. Good aerial photogrpahy, b. Very high resolution satellite imagery (VHR), c. Stereo VHR imagery, d. Pictometry, e. Multi angle VHR imagery, f. Lidar.

Imagery used to derive built up classes that could be considered for global exposure:
Aster – For generation of DEM and urban cover (i.e. 100 City project)
Landsat – Great asset, globally avaialble. Yet, no global urban layer derived yet. Maybe becuase the difficulty to obtain consistent measure of built -up
MODIS – recently developed global urban cover
NOAA-Night lights. The data have been used to map population exposure. Is it too coarse to asses physical exposure?

2. What information extraction techniques?
What information techniques have been used or could be used to extract exposure information from EO data?
For mapping 2D
a. Visiual analysis (Digitizing) – Typically for mapping building footprints (urban land uses).
b. Collaborative visual analysis (Digitizing) proven very effective in post disaster mapping but validation remains an issue not yet addressed – Typically for mapping building footprints (or urban land uses).

c. Object Oriented classification – set of rules may not be exportable to other built environments – Used for mapping urban land uses but it is considered also for mapping buildng footprints .

– Different strategies for extraction of 3D information from stereo:
a. automatic extracton of DSM.
b. semiautomatic extraction of building height (building footprint required).
c. automatic extraction of building height (research in progress)
d. automatic extraction of block hieght information

For qualifying exposure
a. In urban land uses, the mapping includes also the qualification of the class (i.e. slum mapping). Object Oriented classification for slum mapping

b. The building stock may be qualified after having mapped the footprints or volume. Qualification of footprint or building volume can be obtained using
b.1 multi-angle VHR imagery
b2. Pictometry (for the moment not really an option because unavailable for large part of the world except the US)
c. Size of buildings, geogrpahical setting, spatial arrangement of buildings, presence of vegetation to name a few.

Field work
Field data are essential for the the qualification of the built-up area or building stock. For example, typology of buildng can not be determined from EO data alone.

3. What exposure informaiton?
What exposure information should be generated for the disaster risk community
3.1 Exposed asset and its locatioin –
a. Building stock and other information from cadaster
b. Footprints form VHR imagery or aerial photography
c. Cadaster information summarized at aeral unit (block, census, other) level as shape file
d. Homogeneous human patches (as in slums)
e. Built up stratified as urban land uses classes
f. Soil-sealing map (as proxi for built up and thus exposure)

3.2 Attribute of exposed asset
a. Quality of construction: Building height, construction type,
b. Use of building: residential, public, recreational, …

3.3 Exposed assets other than physical
Social vulnerability (based on quickbird and Resourcesat 5.6m )

3.4 Global built up
Built up map (Global) 500 m from MODIS (as proxi for exposure)

4. Other issues
4.1 Training
Training of professionals in communities to absorb technology is fundamental
4.2 Governance