Geoscience Reference
In-Depth Information
board the space shuttle in February 2000. SRTM was used to produce a digital elevation
model (DEM) from 56Sto60 N at 3-arc-second (*90 m) spatial resolution. Average
global height accuracies vary between 5 and 9 m (Farr et al.
2007
) with pixel-to-pixel
noise of *6 m (Rodriguez et al.
2006
), and this is problematic given typical flood
amplitudes. The vertical error has been shown to be correlated with topographic relief with
large errors and data voids over high-relief terrain, whilst in the low-relief sites, such as
river valleys, floodplains and wetlands, errors are smaller (Falorni et al.
2005
). However,
despite better accuracy over low-relief terrain, pixel-to-pixel noise is not reduced and the
X- and C-band radars used for the SRTM mission only partially penetrate vegetation
canopies such that for forested floodplains, the DEM is corrupted by vegetation artefacts.
Accordingly, the SRTM spatial resolution cannot capture the floodplain and wetland
microtopography that can be critical to an understanding of flow dynamics (see Trigg et al.
2012
) and at their native resolution have noise that can be larger than the flood ''signal''.
Attempts at solving these problems by post-processing to remove the vegetation signal and
spatial averaging to reduce uncorrelated noise (Paz et al.
2010
; Paiva et al.
2011
,
2013
)
have been attempted with limited success, and with careful handling, SRTM data have
been shown to be useful for some flood modelling problems (Sanders
2007
; Wilson et al.
2007
; Di Baldassarre et al.
2009
; Neal et al.
2012
).
Other global terrain data sets include the ASTER GDEM (global digital elevation
model), the SPOT 5 DEM and the forthcoming TanDEM-X products. ASTER GDEM is a
30-m spatial resolution DEM developed using stereo-photogrammetry and available from
83Sto83 N. However, its accuracy of 17 m at the 95 % confidence level (Tachikawa
et al.
2011
) means that SRTM has significant advantages for most flood modelling studies.
More promising perhaps is the TanDEM-X global DEM available from 2014 which will
use X-band synthetic aperture radar interferometry to create a global DEM with *12-m
spatial resolution and target accuracy of better than 2 m. Whilst potentially of greater
accuracy and resolution than SRTM, the use of X-band radars will mean that TanDEM-X
may still be corrupted by vegetation artefacts that may be difficult to fully remove even
with sophisticated processing techniques.
2.2 Remote Measurements of Inundation Extent
Globally available remote measurements of inundation extent are reviewed in detail by
Marcus and Fonstad (
2008
) and Schumann et al. (
2012
), and these are made principally
using (a) optical sensors; (b) passive microwave instruments; or (c) synthetic aperture
radars. Visible-band satellite imagery (e.g., 30-m resolution Landsat or coarser 250-m
resolution MODIS data) can detect floods (e.g., Bates et al.
1997
); however, cloud cover
and restriction to daytime only operation may limit the utility of these data. Passive
microwave instruments, such as the scanning multichannel microwave radiometer
(SMMR), have good temporal but limited spatial resolution (6-day revisit time and 0.25
pixels in the case of SMMR) that limits their use to particular types of study (see, for
example, Hamilton et al.
2002
). For these reasons, SAR data are often preferred for flood
remote sensing.
SARs are active systems that emit microwave pulses at an oblique angle towards the
target. Open water acts as a specular reflector, and the microwave energy is reflected away
from the sensor so such areas appear as smooth areas of low backscatter in the resulting
imagery. Terrestrial land surfaces, by contrast, reflect the energy in many directions,
including back towards the sensor, and therefore appear as noisy high-backscatter zones.
These differences allow flood extent to be mapped using a variety of techniques to an
