IEEE Geoscience and Remote Sensing Magazine - June 2019 - 58

STEP 1: CONTINUUM REMOVAL
The depth of the spectral features (and, as a consequence,
the definition of the deepest feature) is measured in respect
to the spectral continuum calculated for the spectra. This
can be done based on either a convex hull quotient (multiplicative) or a hull difference (additive). Also, for the Wavelength Mapper, we recommend using the hull quotient as
the default continuum-removal technique, although hull
difference is available for specialized applications, such as
absorbance spectra.

Continuum-Removed
Reflectance (Offset for Clarity)

STEPS 1 AND 2: SPECTRAL SUBSET
In both processing steps, a spectral subset can be defined to
influence the results. The step 1 spectral subset establishes
the search range in which the minimum feature can occur.
Features outside that range will not be detected, which allows the user to exclude, e.g., water absorption bands in the
data. The step 2 spectral subset delineates how the detected
minimum features from step 1 are mapped onto a rainbow
color table (Figure 5), which ranges from blue (the minimum wavelength in the step 2 range) to magenta (the maximum wavelength in the step 2 range). In the following, we
present a few typical combinations that can be used.

2.1 2.15 2.2 2.25 2.3 2.35 2.4
Wavelength (µm)
Zoom Ramp
Overview Ramp
Alunite
Muscovite (Short)
Muscovite (Long)

Hornblende
Chlorite
Chlorite + Muscovite

FIGURE 5. A continuum-removed example of spectra of Al-OHdominated mineral features (e.g., alunite and two compositions of
muscovite) and iron/Mg-OH-dominated minerals (e.g., hornblende
and chlorite). The two color ramps demonstrate the ranges for a
typical SWIR overview map (2.1-2.4 μm) and a spectral zoom on
the Al-OH minerals. The vertical dashes highlight the minimum
wavelength positions of the deepest feature. For chlorite + muscovite, the muscovite feature is highlighted as well (see the "Step 1:
Zoom" section). The y-axis tick interval is 0.05 reflectance units.

SUBSET STEPS 1 AND 2 INCLUDE THE
SAME LARGE SPECTRAL RANGE
This is the most common default scenario. The deepest features are detected and color-mapped in the same spectral
range. An example would be a SWIR-active mineralogy
overview map in which the entire 2.1-2.4-µm SWIR range
is used for step 1 and step 2. All pixels will receive a hue
from the rainbow color table, ranging from blue (2.1 µm)
to magenta (2.4 µm). This will create a quick overview of
the different mineral groups, resulting in hues of greenishblue for minerals with Al-OH features and reddish hues for
minerals with iron/Mg-OH features (the overview ramp in
Figure 5). However, within these groups, the hues are quite
similar, and small shifts in wavelengths are not immediately discernible in all cases.
STEP 2: ZOOM
Because determining the minima of the features is considerably more time consuming than producing the color
outputs, it can be beneficial to start with a wide step 1
range and use the step 2 subset to zoom in on a particular
wavelength range to achieve better color separation in that
zoomed section of the spectrum. An example would be to
run step 1 on the 2.1-2.4-µm range and step 2 on the 2.15-
2.25-µm range to achieve distinct color separation of, e.g.,
alunite (roughly 2.18 µm) from muscovite (approximately
2.2 µm) (the zoom ramp in Figure 5). Any pixels with a
dominating feature outside the step 2 subset will show up
in gray levels (e.g., chlorite in Figure 5), while areas having
absorption features in the step 2 range are distributed over
the different parts of the rainbow color table.
STEP 1: ZOOM
In some instances, it may be beneficial to return to step 1
to perform a spectral zoom to focus the search for the local minimum on a more restricted wavelength range. An
example could be an area that is rich in minerals with iron/
Mg-OH features (e.g., chlorite) but that also contains some
muscovite with Al-OH features. The bottom spectrum in
Figure 5 shows an example of a mixture of chlorite with
some muscovite. If we intend to map the changes in muscovite chemistry through wavelength shifts in the Al-OH
feature, we need to previously focus the step 1 subset on
a spectral range near 2.2 µm. Failure to focus on the 2.2-µm
subset will result in the detection of mineral features in
the 2.3-µm part of the spectrum because they may be
more dominant in the mixture (see the vertical lines
in Figure 5 for the position of the chlorite and muscovite
features, respectively). This situation would result in pixels
being displayed in gray tones, even though they could contain a measurable muscovite fraction.
STEP 2: DEPTH STRETCH
During step 2, the feature depth information is used as the
intensity scale for the output colors. Choosing the correct
depth stretch values guarantees good intensity contrast,
IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE

JUNE 2019

IEEE Geoscience and Remote Sensing Magazine - June 2019

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - June 2019