Hyperspectral Imagery in Fluvial Environments - Fluvial Remote Sensing for Science and Management - page 75

classified field measurements as 'ground truth' (Legleiter
et al., 2002; Marcus, 2002). Perhaps a better approach in
some cases is to remove ground classification from the
analysis and replace it with groundmeasurement of actual
physical quantities (water depth, substrate size, vegeta-
tion cover percent, etc.). A field spectrometer is one way
of directly comparing hyperspectral imagery with in situ
optical properties of the water, while at the same time
aiding in atmospheric correction (Figure 4.4). Another

(a)

(b)

Figure 4.3 Scanning errors due to complex platform motions:
(a) the helicopter platform, and (b) the resulting imagery before
geometric correction.

the 'truth' and comparing this with the digital image esti-
mates; this is often called 'ground truthing'. The issue is
that in some cases, the hyperspectral image analyses can
produce habitat maps and estimates of physical, biologi-
cal, and chemical parameters that are so good, that they
are actually better than the field measurements. This is
particularly the case in areas of high spatial complexity
and mixing. Therefore, it is often not correct to use the

Figure 4.4 Ground validation through in situ measurement of
river optics using a field spectrometer. Such an approach can be
used to compare the spectra of 'pure' stream classes to one
another and to hyperspectral imagery, and it can also be used to
aid in atmospheric correction.