EY
ov.tr)
on,
life and
al post-
spatial-
vertical
Pre and
al level.
improve
ilti level
sensing
itions in
damage
damage
; most of
to assist
es are a
despread
he major
ulnerable
ure and
ptions of
coming
detailed
allenging
d relief
sult made
quake, as
normally
for post-
valuable
ssessment
1t (Turker
Ithough a
the most
mitations,
|. imagery,
rs. Visual
ire, 2001)
flexible
ly sensed
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004
data are useful for damage assessment in urban areas, but each
sensor has its own characteristics in terms of viewing direction,
and spatial and temporal resolution. Oblique viewing
characteristics, and high temporal and spatial resolution are the
main requirements for data used in damage detection. Although
satellite imagery provides synoptic coverage, cloud coverage,
high cost of high spatial resolution data and vertical viewing are
the main disadvantages of satellite imagery. The characteristics
of urban areas, 3D nature of building damage and rapid onset of
the disaster itself create the main obstacles for data selection.
Even using high spatial resolution data may not help to detect
pancake collapse with intact roofs. The use of night time
imagery to detect damaged areas can speed up data gathering
due to high revisit time, e.g. captured by DMSP/OLS. But it
suffers from low spatial resolution (2.7 km; EDM, 2000). Laser
scanners have significant advantages as they can give
information about building heights and allow the creation of
DSMs. On the other hand, the high cost, complexity in
processing the data and occlusion are the main disadvantages of
the technique. Standard aerial photography can provide high
spatial resolution for urban areas, but the cost of data
acquisition and time requirement for processing are the main
obstacle.
Aerial video imagery is useful for rapid data acquisition at
lower cost. Moreover, the oblique viewing characteristic allows
imaging of building façades, which can improve damage
assessment. However, the low quality of the video imagery is a
major problem in digital analysis. Mitomi et al. (2000) used
aerial video imagery to detect damaged areas following the
1995 Chi-Chi (Taiwan) and 1999 Kocaeli (Turkey)
earthquakes. In conclusion, even though none of the sensors
alone can fulfil all requirements, limitations can be overcome
by integration of different data sources. In the case of 11
September 2001 attack on the World Trade Center, a LIDAR
sensor, high resolution digital camera, and thermal camera were
used to assess damaged areas. Digital images were used to get
information about rubble piles, while LIDAR data were used to
calculate rubble volumes and thermal image provided helpful
information for identification of fires (Rodarmel et al., 2002).
This research aims to integrate, in terms of synergy, space-
borne and airborne imagery to improve damage assessment.
The synoptic coverage of satellite imagery can provide
information at the regional level. Oblique airborne video
imagery can complement the satellite imagery analysis at the
local level. On the other hand, despite of usefulness of remote
sensing technology, the use of this technology is still not
widespread in emergency management activities. Limited
awareness among emergency managers of remote sensing
technology and its potential for disaster management, and
frequent lack of full understanding of emergency activities
among remote sensing technologist are the main reason for this
situation (Bruzewicz and McKim, 1995). Information is only
useful when it is accessible, understandable and manageable by
the user (Ayanz et al., 1997). Therefore, this research also aims
to define the end user information requirements at different
levels in emergency activities to assess the effectiveness of the
research results.
1.3 Case Study: 1999 Kocaeli Earthquake, Turkey
Kocaeli is situated in the north-western part of Turkey, lying
within the North Anatolian Fault Zone, which is one of the
world's longest and best known faults (Figure 1).
Industrialization and high population density are the major
687
characteristics of the region. A devastating earthquake (M: 7.4)
occurred on 17 August 1999 and resulted in widespread and
extensive damage, affecting 4 provinces in the region:
Adapazari, Kocaeli, Bolu, Yalova. Over 15.000 people are
estimated to have died and 40.000 building collapsed or were
heavily damaged after the earthquake. Golcuk was one of the
most damaged towns in the region with a death toll of 5.384 and
2.300 collapsed building. Because of the effect on the large
area, assessing the damage for the relief works was a
challenging task in the time of emergency after the disaster.
Pancake collapse was the most serious damage, as there was the
least opportunity for people to escape.
Locatlon of August 17, 1999 Turkish Earthquake
1992
-
i
19039 2 i
a | } ! ES | Black Sea | -4
Sd eit iL s
Me rae thee
/ Em rm BE E NIU s
—— “ogre inptollen Fault 70v,
7 ment" "74 70 wet Bie. SAR M
Ye
997 m Ankara ac
Epicenter 0 100. k &
> site TURKEY e
5 Hiatorical earthquake epicenter and magnitude
»-- Extent of surface rupture
==> Directiona of relative motion on fault
Figure 1. North Anatolian Fault Zone and case study area
(http://geohazards.cr.usgs.gov accessed on 14 November 2003)
2. DATA AND METHODOLOGY
In this research two different types of remote sensing data were
used to detect damaged areas: moderate resolution satellite
imagery and aerial video imagery. For regional assessment of
damage, pre- and post-earthquake (15 July and 20 August 1999,
respectively) SPOT 4 HRVIR panchromatic and multispectral
images were used. For local level damage assessment, aerial
video imagery taken by a media agency on the day of the
earthquake (17 August 1999) was used in the research. Total
footage was around five minutes with a resolution of 720x576
lines. Damage assessment surveys results at the regional level
were provided by the General Directorate of Disaster Affairs,
Turkey. At the local level, a damage assessment survey, carried
out by the Architecture Institute of Japan (2000) for Golcuk
city, was used to verify the results of the analysis. Cadastral
boundaries provided by Golcuk Municipality were used to
aggregate damage information at the local level. The research
was carried out in four steps: (i) analysis of user information
requirements of emergency agencies in Turkey, (ii) analysis of
Spot imagery, (iii) analysis of aerial video imagery and (iv)
evaluation of results by comparison with the ground survey.
An assessment of user information requirements was carried out
by interviewing 14 key informants from different emergency
organizations in Turkey, such as national, regional and local
government agencies, NGOs and a research centre participating
in the emergency activities in Turkey. Institutional background,
data requirement and previous experiences were investigated in
the interviews.
Spot imagery was analysed using change detection methods
shown in Figure 2. After geometric correction, a critical step in
