Geoscience Reference
In-Depth Information
Earthquakes and other natural hazards can cause disasters of uncontrollable
magnitude when they hit large urban areas. Emergency response and early recovery
assessment in earthquakes require rapid and reliable damage assessment and loss es-
timation. In the case of suddenly occurring earthquakes, remote sensing data can be
reliably used to create fast draft damage maps of the affected urban areas which pro-
vide valuable information to support emergency response teams and decision making
during the recovery process. Remotely sensed images ranging from very high resolu-
tion to medium resolution have been widely used to derive information and estimation
for damage assessment (Bendimerad, 2001; Kaya et al., 2005; Pal et al., 2006; Sertel
et al., 2007; Stramondo et al., 2006; Turker and San, 2003). Moreover, multitemporal
remote sensing data can serve as a basic data set to support post-disaster planning.
Different remote sensing methods have been used by many scientists to identify
earthquake-induced damage areas. Sertel et al. (2007) investigated the relationship
between semi-variogram metrics and degree of earthquake damage using transects
over an earthquake area. Turker and San (2003) used differences between merged
pre- and post-event SPOT high resolution visible (HRV) data to reveal the location
of earthquake-induced changes. Stramondo et al. (2006) used coherence and correla-
tion maps from Advanced SAR (ASAR) and change maps from advanced spaceborne
thermal emission and refl ection radiometer (ASTER) to analyze the capabilities and
limitations of satellite remote sensing to detect damage due to earthquakes. Kaya et al.
(2005) used government statistics and SPOT HRV data to estimate the proportion of
collapsed buildings in an earthquake area. Although there have been a lot of earth-
quake damage assessment studies using different remote sensing methods, there has
not been that much research on the application of Fourier transform to satellite image
for an earthquake case. This research focuses on integrated usage of Fourier transform
and level slicing to identify earthquake induced damage areas, also detailed accuracy
assessment of proposed method was conducted using 1/5,000 scale damage map data
and error matrix analyses.
Fourier transforms have been applied to different remote sensing applications.
Lillo-Saavedra et al. (2005) used Fourier transforms to fuse panchromatic and multi-
spectral data obtained from Landsat ETM+ sensor. Westra and De Wulf (2007) used
Fourier analysis of Moderate Resolution Image Spectrometer (MODIS) time series
data to monitor the fl ooding extent. Pal et al. (2006) used FFT fi lter to extract linear
and anomalous patterns. Their results showed that numerous lineaments and drainage
patterns could be identifi ed and demarcated by FFT fi lters.
In this study, the following steps were conducted to accurately identify the location
and magnitude of earthquake induced damages in an urban area and to quantify the
accuracy of the proposed method: (i) pre- and post-earthquake images of the region
were geometrically and atmospherically corrected, (ii) The FFT was applied to pre-
and post-earthquake images and images were fi ltered in the frequency domain, (iii) a
difference image was generated using IFFT-pre- and post-earthquake data, (iv) level
slicing method was applied to difference image to identify the earthquake-induced
damages, (v) accuracy assessment was performed by comparing the results of the pro-
posed method with the 1/5,000 scale damage map of the earthquake area.
