Farm Animal Metabolism and Nutrition

case of NIRS) involved in finding a calibra-
tion equation for the constituent of interest.
The field of using multivariate statistical
procedures for the quantitative and qualita-
tive analysis has come to be known as
chemometrics, and is an integral part of
utilizing NIRS.
Since the constituent of interest is both
the analyte being determined and part of
the media or matrix, and since all the other
constituents in the sample are most
probably also changing in relative amounts
from sample to sample, it is extremely
important that the proper samples be used
when developing a calibration. Because all
the organic constituents in a feed sample
are likely to absorb NIR radiation at all
wavelengths to at least some degree, and
because all or most of the constituents will
vary simultaneously in relative amounts
present, one determines the values for a
given constituent not only in terms of
itself, but also in terms of everything else
present. As a result, the structure of a data
set is probably the most important part of
developing and utilizing NIRS, and the

misunderstanding of this point probably

leads to more problems and disillusion-

ment with NIRS than anything else.

Sample sets and calibration development

In using a single wavelength, it is easy to

see that anything (other than changes in

the analyte) that might alter the absorbance

will result in some error in the determina-

tion of the analyte. For example, in Fig.

9.4, if spectrum A were to be shifted to the

right to give the results in B (temperature

changes can easily cause such peak shifts),

then the absorbance value read at the

wavelength where the peak should be

would result in an incorrect value for the

analyte. In Fig. 9.4C and D, we see the

effects of two kinds of baseline shift. In Fig.

9.4C, the entire spectrum has shifted

upwards (this is very common in NIR spec-

tra due to particle size differences between

samples), resulting in higher absorbance

values, although a simple baseline correc-

tion can be made. In Fig. 9.4D, the baseline

188 J.B. Reeves III

Absorbance

1000 1200 1400 1600 1800 2000 2200 2400

Wavelength (nm)

High-quality lucerne hay

Wheat straw

Fig. 9.2.NIR spectra of high-quality lucerne hay and wheat straw.