Geography Reference
In-Depth Information
Such sensors are called 'Hyperspectral' and can have
hundreds or even thousands of bands with resolutions as
small as 0 . 002 μm. Whilst such hyperspectral sensors have
huge potential, their usage in river sciences has been rel-
atively limited and most of the progress in fluvial remote
sensing rests on standard colour imagery with the con-
ventional three bands of R ed, G reen and B lue(hencethe
term RGB imagery) which equates to a relatively coarse
spectral resolution of approximately 0
memory. When radiation reaches a device, the intensity
of radiation must be converted to some proportional
brightness scale which can then be represented on an
image. In the case of digital devices, this proportional
brightness is termed the Digital Number (DN). The digi-
tal number is the dimensionless actual value of the pixel
that can be seen if the image is accessed with image pro-
cessing software. Typically, these pixel values are scaled
to increasing powers of 2. For example, standard RGB
imagery contains three bands, each of which has pixel val-
ues ranging from 0 to 255. These 256 possible values arise
from data storage in an '8 bit' binary format meaning that
each DN value is coded with 8 binary digits with possible
values of 0 or 1 thus leading to 2 8 (256) possible values for
the image pixels. However, more advanced sensors and
satellites will frequently use higher 'bit-depths' of 11 or
12 bits thus leading to a wider range of 2048 (2 11 )oreven
4096 (2 12 ) DN values. This higher number of DN values
can help in resolving finer differences in image brightness.
In river sciences, radiometric resolution can be an impor-
tant parameter when trying to measure river properties
through the water interface (Legleiter et al., 2009).
In summary, from the point of view of an end-user, the
fundamental properties of a remote sensing data acqui-
sition system can be described by four key parameters:
Spatial resolution, spectral resolution, temporal resolu-
tion and radiometric resolution. Spatial resolution is
often considered as the primary parameter as it defines
the size of the smallest object which can be resolved on
the ground. Spectral resolution can be crucial in iden-
tifying certain materials, such as chlorophyll, based on
their reflection of light as a function of the wavelength
of the incident light. Temporal resolution is obviously
crucial in change detection studies. Finally, radiometric
resolution, often called 'bit-depth', defines the amount of
information devoted to the storage of each image pixel.
Higher radiometric resolutions allow for the recording of
smaller differences in image brightness.
m.
One key advantage of widely available colour imagery
is its very high spatial resolution. One of the most
fundamental descriptors of remote sensing data, spatial
resolution refers to the ground footprint of a single image
pixel on real ground. This distance is generally quoted
as a linear unit with the underlying assumption that the
pixels are square. The spatial resolution of a dataset will
define the smallest object that can be identified. Whilst
there is no absolute rule for the number of pixels required
to define a simple object (e.g. a boulder), our experience
has shown that a minimum of 5X5 pixels are required
in order to get an approximation of the object shape
whilst 3X3, or even 2X2, pixels are required to establish
to presence of an object of undefined shape in the image.
In parallel with spatial resolution, temporal resolution
refers to the elapsed time between repeated imagery.
Repeated image sampling has been somewhat less
exploited in fluvial remote sensing. While studies of large
rivers based on satellite imagery have been able to exploit
the regular revisit frequency of orbital sensors (Sun et al.,
2009; Frankl et al., 2011), airborne data is not acquired
with the same regularity and studies reporting change
based on airborne data are much less frequent. As a
result, substantial progress remains to be made in terms
of monitoring rivers and examining changes occurring at
the smaller spatial resolutions that can be detected with
airborne remote sensing. However, repeated imagery,
including video imagery, has been successfully used
at smaller scales for laboratory studies (see Chapter
13) and reach based studies (see Chapters 15 and 16).
Furthermore, a largely un-exploited archive or terrestrial
and airborne archival imagery exists for many parts of
the world which does indeed include riverine areas. If
issues such as image georeferencing (spatial positioning
of the imagery), and image quality can be addressed
(see Chapter 8), then these images could provide a very
important source of data sometimes dating as far back as
the nineteenth century.
The final parameter, radiometric resolution is easily
confused with spectral resolution. Here the term 'radio-
metric' refers to the recording of data in the sensors
.
2
μ
1.2.2 Ashort introductionto 'river friendly'
sensorsandplatforms
A remote sensing 'platform' is simply the physical support
which carries the 'sensor' that does the actual data collec-
tion. We have illustrated four classic and new platforms in
Figure 1.1. This distinction between platform and sensor
is not always clear, especially in the field of satellite remote
sensing. For example, the TERRA satellite platform car-
ries both the MODIS and ASTER sensor. However, the
commercial term 'QuickBird' is used to describe both
 
Search WWH ::




Custom Search