412
purposes, if samples need to be stored before measurement, they should be kept
cool, moist, and dark and measured as soon as possible. It is good practice to keep
the intensity of the light source at the minimum needed for a good signal-to-noise
ratio, because leaves can get burned by the lamp. Defining a threshold for a “good”
measurement, for example, based on the reflectance at the so-called NIR shoulder,
which is the highest point of the reflectance curve at the beginning of the NIR, can
be helpful. Spectral measurements should be made under dry conditions; instruments can be damaged by water, and water films alter the spectral characteristics
of leaves.
Measurement protocols should specify how many leaves per plant to measure
and which leaves to select. Generally, this means clarifying if a study deals with
“ideal” or “average” plant individuals and if an entire plant should be characterized
or only certain layers, such as the top canopy. Measuring mature, healthy, sunlit
leaves is a good strategy for characterizing species or functional groups. For studies
dealing with disease detection, asymptomatic and symptomatic leaves should be
measured, ideally at different stages of the disease. Selecting leaves at random,
including all ages, canopy layers, and stress levels, can make sense when aiming to
relate leaf chemistry to ecosystem processes or when scaling leaf-level chemistry to
plot-level estimates. If the aim is to characterize entire individuals, leaves from all
canopy layers can be included, with the number of measurements per layer reflecting plant size, growth form, and architectural complexity. However, when leaf-level
spectra are being matched to spectra acquired with RS, it makes sense to select only
leaves from the layer that is captured by the sensor (i.e., from the top of the canopy).
If possible, measuring the midvein should be avoided. Measuring different spots on
the same leaf is also usually unnecessary, at least for small- to medium-sized leaves,
because spectral variation at the leaf level is generally small. However, it is important to check the quality of every spectrum. Ideally, quality checks are done immediately after each measurement; bad measurement can be flagged for subsequent
filtering, which considerably reduces preprocessing time.
15.3.2 Proximal Canopy-Level Spectroscopy
Proximal canopy spectra can be sampled with handheld spectrometers, robotic systems, and UASs. One important differentiation is between nonimaging and imaging
systems. Nonimaging spectrometers integrate over a defined amount of time the
spectral reflectance of an illuminated area; the output is one spectral curve per measurement. Instruments used for leaf-level spectroscopy (Sect. 15.3.1) fall into this
category. The same instruments can be used in a handheld mode or mounted on a
platform to sample spectra at the proximal canopy level, and the reach of instruments can be expanded using long fiber-optic cables attached to a beam. Creative
solutions include mounting a spectrometer on a bike and using it as a mobile platform (see “reflectomobile” in Milton et al. 2009). Imaging spectrometers sample
spectra in a spatially resolved fashion. The collected data are commonly represented
A. K. Schweiger
purposes, if samples need to be stored before measurement, they should be kept
cool, moist, and dark and measured as soon as possible. It is good practice to keep
the intensity of the light source at the minimum needed for a good signal-to-noise
ratio, because leaves can get burned by the lamp. Defining a threshold for a “good”
measurement, for example, based on the reflectance at the so-called NIR shoulder,
which is the highest point of the reflectance curve at the beginning of the NIR, can
be helpful. Spectral measurements should be made under dry conditions; instruments can be damaged by water, and water films alter the spectral characteristics
of leaves.
Measurement protocols should specify how many leaves per plant to measure
and which leaves to select. Generally, this means clarifying if a study deals with
“ideal” or “average” plant individuals and if an entire plant should be characterized
or only certain layers, such as the top canopy. Measuring mature, healthy, sunlit
leaves is a good strategy for characterizing species or functional groups. For studies
dealing with disease detection, asymptomatic and symptomatic leaves should be
measured, ideally at different stages of the disease. Selecting leaves at random,
including all ages, canopy layers, and stress levels, can make sense when aiming to
relate leaf chemistry to ecosystem processes or when scaling leaf-level chemistry to
plot-level estimates. If the aim is to characterize entire individuals, leaves from all
canopy layers can be included, with the number of measurements per layer reflecting plant size, growth form, and architectural complexity. However, when leaf-level
spectra are being matched to spectra acquired with RS, it makes sense to select only
leaves from the layer that is captured by the sensor (i.e., from the top of the canopy).
If possible, measuring the midvein should be avoided. Measuring different spots on
the same leaf is also usually unnecessary, at least for small- to medium-sized leaves,
because spectral variation at the leaf level is generally small. However, it is important to check the quality of every spectrum. Ideally, quality checks are done immediately after each measurement; bad measurement can be flagged for subsequent
filtering, which considerably reduces preprocessing time.
15.3.2 Proximal Canopy-Level Spectroscopy
Proximal canopy spectra can be sampled with handheld spectrometers, robotic systems, and UASs. One important differentiation is between nonimaging and imaging
systems. Nonimaging spectrometers integrate over a defined amount of time the
spectral reflectance of an illuminated area; the output is one spectral curve per measurement. Instruments used for leaf-level spectroscopy (Sect. 15.3.1) fall into this
category. The same instruments can be used in a handheld mode or mounted on a
platform to sample spectra at the proximal canopy level, and the reach of instruments can be expanded using long fiber-optic cables attached to a beam. Creative
solutions include mounting a spectrometer on a bike and using it as a mobile platform (see “reflectomobile” in Milton et al. 2009). Imaging spectrometers sample
spectra in a spatially resolved fashion. The collected data are commonly represented
A. K. Schweiger