Tải bản đầy đủ - 0trang
II. Instrumentation for Reflectance Measurements
MARION F. BAUMGARDNER ET AL.
Normal to sample
FIG. 1. Geometric parameters describing reflection from a surface: 0, zenith angle; 4,
azimuth angle; o,beam solid angle; a prime on a symbol refers to viewing (reflected)conditions.
Figure 1 illustrates the basic geometric relationships between incoming
radiation and outgoing radiation using the previously described terminology.
The reflecting properties of a surface are most precisely described using a
parameter called the bidirectional rejectance distribution function (BRDF).
The defining equation for the BRDF is
The angles are shown in Fig. 1. The BRDF is the ratio of a radiance to an
irradiance; therefore, it has the units of sr-'. If the numerator and denominator of the expression are spectral quantities, then a spectral BRDF has been
defined and is usually denoted by the symbol A. A careful examination of Fig.
1 reveals that the BRDF is the ratio of two differential solid angles. This is a
mathematical abstraction that is closely realized by many physical situations
in which the incident and reflective solid angles are small enough to
approximate the differential case. The physically measured BRDF is therefore an average fr value over the parameter intervals. The incident and
reflected solid angles, however, need to be small to obtain a good estimate of
the true BRDF.
The measurement of the BRDF is, however, a particularly difficult
problem. It would be necessary to place a sensor at the surface to measure the
incoming radiation and then take that sensor, or another sensor, and place it
in the viewing position necessary to measure the reflected radiation. Although this represents a possible approach, an experimentally more convenient method uses a reflectance standard in the measurement procedure. The
REFLECTANCE PROPERTIES OF SOIL
FIG.2. Illustration of procedure for measuring the reflectance factor. The response of the
sensor to a perfectly diffuse, ideal reference is recorded, and then the response of the sensor to a
target of interest is recorded under the same illumination conditions.
geometry of this method is illustrated in Fig. 2. In this method a single sensor
located in the viewing position is used to view the reflected radiation from a
perfectly diffuse ideal reflector as well as that from the scene of interest. If the
scene and the perfectly diffuse ideal reflecting surface are viewed and
illuminated under identical conditions, the ratio of the two measurements is
referred to as the reflectance factor of the scene (Nicodemus et al., 1977).
If the geometry of the situation resembles that of Fig. 1, then the quantity
we are measuring is referred to as the bidirectional reflectance factor (BRF). If
differential solid angles are not assumed and real conical solid angles are the
case, then the quantity being measured is called the biconical reflectance
factor. If the conical solid angles for both incident and reflected radiation
include all directions, i.e., the hemisphere, then the measurement is referred to
as the bihemispherical rejectance factor. Also, the measurement may be
referred to as the directional-hemispherical reflectance factor, hemisphericalconical reflectance factor, etc., depending on the instrumentation setup. In
many real measurement situations the magnitudes of the solid angles
involved are small enough so as to approximate the differential situation. For
this reason, the results of the reflectance factor measurement are referred to as
the bidirectional reflectance factor, whereas in reality they are actually the
results of a biconical reflectance factor measurement.
The principal advantage of the reflectance factor method of measurement
is that the sensor can be kept in its viewing position and it is only necessary to
MARION F. BAUMGARDNER ET AL.
measure the radiation from the reflectance standard and that from the scene
of interest within a short time period and then take the ratio to obtain the
desired reflectance factor measurement. The method is amenable to both
laboratory and field measurement situations.
An important part of the measurement procedure just discussed is the
reflectance standard. A perfectly diffuse reflecting surface is one that reflects
equally in all directions. An ideal reflector is a surface in which all of the
energy falling on the surface is reflected. Again, this is an abstraction, but
physical surfaces have been prepared which approximate this ideal situation.
Examples of some of the reflecting materials and coatings used for diffuse
surfaces include magnesium oxide, barium sulfate (Grum and Luckey, 1968;
Billmeyer et al., 1971; Shai and Schutt, 1971; Young et al., 1980), homopolymers and copolymers with fluorine substitution (Schutt et al., 1981),
and canvas panels (Robinson and Biehl, 1979). Barium sulfate surfaces have
received the widest acceptance.
Specially prepared barium sulfate surfaces approximate a perfect diffuser at
angles within 45" of a normal to the surface. The reflectance of properly
prepared barium sulfate surfaces is over 0.9 for most wavelengths in the
reflective spectrum. Departures from perfectly diffuse or ideally reflecting
properties of the reflectance standard can be documented and accounted for
during the analysis of the measurement data (Robinson and Biehl, 1979).
Primary reflectance factor calibration measurements are usually made with
tablets that are fabricated out of pressed barium sulfate powder. The results
of these measurements are referred to standard data that have been made
available by the manufacturers of this special barium sulfate reflectance
powder. The resulting measurements are then used to calibrate painted
barium sulfate reflectance standard panels. The pressed barium sulfate tablets
can often be directly used in the laboratory, but for field situations 1.25 x
1.25 m painted panels are usually prepared. The panels are covered when not
in use and are otherwise handled very carefully to avoid contamination from
Figure 3 is a schematic diagram of the angles involved in a physically
realizable measurement situation. In this case, the source and the sensor have
real apertures which produce conical solid angles rather than differential
angles as indicated in Fig. 1. In addition, the sensor has a field of view
determined by its internal optics and indicated on the diagram by the angle p.
The amount of reflected power gathered by the sensor is proportional to the
square of the field of view, the sensor aperture area, the irradiance, the
REFLECTANCE PROPERTIES OF SOIL
'sensor field of view
FIG. 3. Schematic diagram of illumination and viewing angles involved in a physically
realizable measurement situation.
irradiance angles, the sensor view angles, and, of course, the bidirectional
reflectance distribution of the target. The angular relationships of the source,
target, and sensor can significantly affect the measured reflectance of the
target. For example, the reflectance measured by a sensor vertically viewing
the soil in a plowed field with deep furrows will be higher when the solar
azimuth angle is parallel with the furrow (no shadows) than when the solar
azimuth is perpendicular to the furrows (shadows present). If the target is
smooth, then the azimuthal orientation of the sensor (or source) is irrelevant.
Usually laboratory samples of a soil are prepared in such a way as to be free
of azimuthal variation. Most field observations are usually made at a sensor
zenith angle of 0" and the data are taken at a variety of solar illumination
angles. Laboratory observations are, again, usually made at a sensor zenith
angle of 0" and an illumination zenith angle of 10 to 30".
As previously stated, the signal power received by the sensor is proportional to the square of the field of view of the sensor. However, this is not the sole
factor in the choice of the particular field of view that is going to be used in a
measurement situation. As shown in Fig, 4, a narrow field of view may not
properly integrate the geometric features of the scene into the signal received
by the sensor. A "wide enough" field of view is necessary in order to represent
properly the geometric features of a target to the sensor. Examples of such
structure are the plant rows in a row crop and the roughness of a plowed field.
However, if the field of view is "too wide," then it is difficult to characterize
properly the sensor zenith angle. In field situations, an appropriate compromise is a field of view of approximately 15". In laboratory situations where
the surface roughness of the target can be controlled, a narrow field of view
such as 1" may be properly used.
MARION F. BAUMGARDNER ET AL.
FIG.4. Illustration of a field of view (FOV) which is large enough to represent properly the
geometric features of the target and a FOV which is too small to integrate the geometricfeatures.
In some cases, the surface roughness of the target may be so extreme as to
make it impossible to characterize properly the reflectance properties with
one measurement. In this case, several observations of the target are made at
a variety of illumination angles or viewing angles to characterize the target
completely. In this way, shadowing effects which might otherwise obscure the
reflectance properties of the target can be taken into consideration. This
method is available only in a low-altitude field instrumentation situation.
Extreme surface roughness may seriously complicate high-altitude observations from aircraft or satellites.
Another factor that enters into the instrumentation geometric problem is
that of the ratio of direct to diffuse illumination. On a very clear, lowhumidity day, most of the incoming illumination (in a field situation) is
directly from the sun. On hazy days, a significant amount of incident
illumination may be indirect due to atmospheric scattering. This is generally
referred to as the diffuse component of the incident illumination. Moreover,
the diffuse component is a function of the wavelength, being proportionally
more at shorter wavelengths. If the diffuse component of incident illumination is a significant part of the total illumination, then it is difficult to
characterize properly the zenith angle of the incident illumination. In a
laboratory situation, the instrumentation system is arranged so as to negate
the effects of diffuse component illumination. In a field situation, it is often
possible to reduce diffuse effects by restricting data acquisition times to those
in which atmospheric conditions do not give rise to extreme diffuse effects. If
it is necessary to take data under hazy atmospheric conditions, then the
diffuse component may be determined by shadowing the target (and standard) from direct illumination and making a reflectance factor measurement.
REFLECTANCE PROPERTIES OF SOIL
This diffuse component is then subtracted from the total illumination in
order to obtain the reflectance factor associated with the direct illumination
Instruments that have been used to measure the reflectance of soil can
generally be divided into two broad classes-spectroradiometers and multiband radiometers. The discussion that follows is a brief summary of the
instrumentation. More detailed information can be found in Robinson and
DeWitt (1983) and Zissis (1979).
Multiband radiometers contain several optical filters to define the spectral
bandpasses. These spectral bandpasses are selected to sample discrete portions of the optical spectrum, e.g., the Landsat multispectral scanner (MSS)
or the thematic mapper (TM) bands. Multiband radiometers can be of two
types-nonchopping (dc) or chopping (ac). The detectors in dc multiband
radiometers are directly coupled to the output amplifiers. The most reliable
dc multiband radiometers are limited to the spectral range of the silicon
detector because detectors such as lead sulfide, which are sensitive beyond
1 pm, are not sufficiently stable.
Many dc multiband radiometers being used to measure the reflectance of
soils in situ contain the Landsat MSS spectral bands or the first four TM
spectral bands. Some companies have built multiband radiometers that make
it relatively easy for researchers to interchange different sets of optical filters.
ac multiband radiometers contain a chopper in front of the detectors to
allow the field of view of the detector to be alternately filled with the internal
reference source and the target. The signal measuring the radiance from the
target is now the ac signal. The dc signal, which is due to the instability of the
detector, is removed. ac multiband radiometers generally use lead sulfide
detectors to measure the radiant power in spectral bands from 1.0 to 3.0 pm,
such as the last three reflective TM bands.
Spectroradiometers.are distinguished from multiband radiometers because
they measure flux in much narrower, continuous spectral bands. Spectroradiometers can also be chopping or nonchopping. Similar to multiband radiometers, the most reliable nonchopping spectroradiometers are limited to the
silicon detector spectral range.
In spectroradiometers, optical dispersion devices replace the simple optical
filters. The dispersion device may be a segmented filter wheel, a circular
variable filter (CVF), a prism, or grating. The differences in dispersion devices
relate to wavelength resolution and spectral scan speed. Circular variable
filter spectroradiometers that operate from 0.4to 2.4 pm have been widely
MARION F. BAUMGARDNER ET AL.
used for laboratory and in situ reflectance measurements. The wavelength
resolution of these CVF instruments is generally 1-2 % of the wavelength.
Spectroradiometers and multiband radiometers may also have an integrating sphere and an artificial light source attached to allow for the measurement of the directional-hemispherical reflectance factor. Many of the
reflectance data cited in the literature are directional-hemispherical reflectance factor measurements.
The systems described above have their own set of advantages and
disadvantages. Multiband radiometers tend to be less costly to buy and
operate. They are lighter and therefore easier to mount on simple pickup
truck booms. Generally, one can obtain a set of multiband radiometer
measurements much faster than a set of spectroradiometer measurements
-fractions of a second compared to 10-120 sec. However, the spectral range
and resolution of multiband radiometers are limited. For example, one can
use reflectance measurements from a multiband radiometer with Landsat
MSS bands to help interpret Landsat MSS data; however, the measurement
could not be easily used to interpret TM data.
Spectroradiometers, on the other hand, lend themselves well to obtaining
sets of reflectance data which can be used to interpret many different sets of
broadband satellite data, e.g., Landsat MSS and TM, the National Oceanic
and Atmospheric Administration’s advanced very-high-resolution radiometer (NOAA AVHRR), and the Nimbus 7 coastal zone color scanner
(CZCS). Spectroradiometers, however, tend to be costly to operate, require
significant time to obtain a set of spectral measurements, and are cumbersome to move in the field. Recent advances in multilinear arrays, however,
may negate some of these disadvantages in the near future.
Instrumentation systems that have been used to measure the reflectance of
soil can be divided into laboratory and in situ or field systems. The laboratory
systems can be further divided into bidirectional reflectance factor systems
and directional-hemispherical reflectance factor systems. Reflectance measurements were initially done in the laboratory using directional-hemispherical spectroradiometer systems. I n situ measurements were made of soil
reflectance as field-worthy spectroradiometer systems were developed during
the late 1960s and early 1970s. With the launch of the Landsat MSS scanners
and the attention given to the Landsat MSS bands, multiband radiometers
became a primary source of field measurements during the late 1970s and
early 1980s. Laboratory measurements of reflectance also continued during
this time. Instrument systems were developed to measure the bidirectional
reflectance factor of soils in the laboratory (DeWitt and Robinson, 1976).
Many instrumentation systems are now available to make reflectance
measurements. However, to utilize and compare measurements from different
systems, researchers need to follow well-defined calibration procedures and
REFLECTANCE PROPERTIES OF SOIL
describe adequately their measurements relative to illumination and viewing
angle, sensor field of view, reference surface, illumination and viewing solid
angles, and target surface preparation. Also, one needs to be cautious about
comparing bidirectional measurements and directional-hemispherical
measurements. Directional-hemispherical measurements are an integration
of bidirectional measurements across all possible viewing angles.
Ill. EFFECTS OF SOIL CONSTITUENTS ON SOIL
With the development of new laboratory and field spectroradiometers
during the past two decades, it is now possible to measure quantitatively the
effects of soil constituents on soil reflectance. One of the early studies to
quantify soil reflectance and to define differences between soil reflectance
spectra was conducted by Condit (1970, 1972). He made directional-hemispherical reflectance factor measurements over the range from 0.32 to 1.0 pm
for 160 soil samples from 36 states in the United States. From these results he
classified the spectral soil curves into three general types (Fig. 5). No attempt
was made to relate quantitatively these spectral properties to other physical
and chemical properties of the soils.
By the mid-1970s laboratory and field instruments were available for
obtaining bidirectional reflectance factor measurements over an extended
range of the reflectance spectrum. Stoner and Baumgardner (1981) and
Stoner et al. (1980a) reported the results of visible and infrared reflectance
(0.52-2.38 pm) measurements for duplicate surface samples of more than 240
soil series obtained from 17 different temperature-moisture regimes in the 48
contiguous states in the United States. From these spectra and a limited
FIG.5. Soil spectral curve types defined by Condit (1970, 1972); reflectance is recorded as
the directional-hemispherical reflectance factor.
MARION F. BAUMGARDNER ET AL.
FIG.6. Soil spectral curve forms defined by Stoner and Baumgardner (1981); reflectance is
recorded as the bidirectional reflectance factor. curve a, Organic dominated; curve b, minimally
altered; curve c, iron affected; curve d, organic affected; curve e, iron dominated.
number of reflectance spectra of soils from other parts of the world, they
derived five soil spectral curve forms (Fig. 6 ) to which they established some
general genetic, physical, and chemical relationships to each spectral curve
form (Table I).
This capability to obtain calibrated data across the visible and infrared
reflectance spectrum provides an important new tool to soil scientists (Cipra
et al., 1971a). The remainder of this section will present the results of
observations made on those soil constituents which account for most of the
variation in soil reflectance.
It is a common observation that most soils appear darker when wet than
when dry. This results from decreased reflectance of incident radiation in the
visible region of the spectrum. Evans (1948), who presented reflectance curves
for three soils in both the wet and dry state, found that the wet samples
showed lower reflectance. Unfortunately, he provided no information about
the soil series or moisture contents of the soils. Brooks (1952) reported 10%
reflectivity of moist Yo10 fine sandy loam over the wavelength range of 0.4 to
2.5 pm but did not indicate the moisture content; for the dry condition, he
reported a reflectivity of 30%. Kojima (1958b) studied the effect of moisture
content on the color of 16 soils. His results, reported in Munsell color
notation, also showed a decrease in reflectance with an increase in moisture.
He made no reference to energy changes related to reflectance and moisture
In “Soil Taxonomy,” the Soil Conservation Service standard for soil
survey over much of the world, the range of change in soil color upon wetting
is given as varying between 1/2 and 3 Munsell color steps (Soil Survey Staff,
1975). No formulas are proposed for predicting change in color between the