Use of Near Infrared Reflectance Spectroscopy 205
physical state can also have spectral effects
similar to those caused by water. Thus
amorphous and molten sugars look very
much like sugars in solution, indicating a
loss in crystallinity as the cause. The
question then remains as to whether these
effects are the reason that NIRS does not
operate as well on high moisture materials.
Efforts to quantify the loss of spectral
information caused by water were carried
out by comparing water-subtracted spectra
with other water-subtracted spectra and
with spectra of amorphous and dry
crystalline materials (Reeves, 1995). Results
showed that spectra for dry crystalline
materials are in general much less similar
than spectra of dissolved or amorphous
materials. However, the loss of information
typically is so large that the question then
becomes how NIRS operates at all on high
moisture samples. Examining the effects of
water on the various compounds by class
showed that polymeric materials, such as
cellulose, starch, pectins, hemicelluloses
and proteins, were the least effected by
water, pH and physical state variations. The
fact that such polymers make up a large
percentage of the total dry matter in either
dry forages or wet silages may account for
why NIRS works reasonably well with high
moisture materials. Thus, for wet samples,
NIRS may only have trouble determining
the composition of the relatively small frac-
tion of the sample comprised of monomeric
materials.
Solid state matrix effects
Another possible reason why NIRS performs
better with high moisture samples than the
work with model compounds would
indicate, may lie in the nature of spectra
found for dry materials. In reality, dry
samples such as forages are complex
mixtures of the various components, and
may contain compartmentalized fractions
and fractions which behave more like solid
solutions than simple mixtures of pure
components. For example, the soluble
monomers in plant leaves are likely to be
compartmentalized either in vacuoles or in
the sap. In samples which have been
crushed or fermented (silages), the sap or
vacuoles may well be mixed with the more
structural parts of the plant tissue. In either
case, one is unlikely to find an isolated
solution of glucose which can crystallize
well.
The question is what happens to
spectra under these circumstances. If a
solution of glucose is dried in the presence
of cellulose, does the spectrum of the
resulting dry mixture look like the
spectrum obtained if dry cellulose and dry
crystalline glucose are simply mixed
together? Results have shown that it is easy
to see the spectrum of glucose in simple
mixtures of cellulose and glucose using
spectral subtraction. However, when a
solution of glucose is added to cellulose
and the mix dried under vacuum while
being stirred continuously, it becomes
difficult to find the glucose. Thus, even for
dry materials, there are spectral inter-
actions which resemble those found with
high moisture mixtures and solutions.
Only further efforts with both dried and
high moisture samples are likely to answer
the question of how the presence of water
in high moisture samples degrades NIRS
results.
Conclusions
While there is still the need for research on
NIRS and the limits to its use, it is a useful
technique. With proper planning and care,
it can reduce the need and time required
for chemical analysis. At present, NIRS
works best on dried materials and requires
constant checking to maintain the accuracy
of calibration equations. How much the
present limits to analysis by NIRS can be
reduced by improvements in instrumenta-
tion, chemometric techniques and wet
chemical analysis procedures remains to be
seen. If the techniques used to determine
fibre were as precise and as non-empirical
as those used for protein, would the
accuracy for fibre determination by NIRS
be as high as that obtained for protein, and
if not why? These and many other