Farm Animal Metabolism and Nutrition

Use of Near Infrared Reflectance Spectroscopy 203

Populations of silage samples often contain
samples of high moisture content and
possess a wide range of moisture contents.
Even though NIR radiation is better able
than MIDIR to handle samples containing a
large amount of water, water still causes
considerable problems. In a sample
containing a large amount of water, the
water bands present are quite large and
tend to overlap and obscure other spectral
features. This can also cause non-linear
responses due to instrumental limitations.
Thus doubling the concentration of a
component does not necessarily cause an
absorption doubling. The end result, as
shown in Table 9.5 (Reeves and Blosser,
1991), is that the results found for high
moisture samples, such as silages, are not
as good as those for the same or similar
samples when dried. The former USDA
NIRS Network and former Pacific Scien-
tific Co. (now FOSS NIRSystems, Inc.)
sponsored studies of this problem which
showed that, no matter how wet silages
were ground or scanned (data not shown),
the results were never as good as for the
same samples when dried. However, as
shown in Table 9.6, NIRS is capable of
measuring many, but not all components of
interest in wet samples (Reeves et al.,
1989). The choice seems either to use NIRS
on dried and, therefore, modified samples
(loss of volatiles such as short-chain fatty
acids and ammonia on drying) or to use
wet unmodified samples and accept the

loss in accuracy. Another alternative is to

use NIR to determine some variables using

dry samples, and other methods, such as

titration, on volatiles.

Many researchers are exploring the

SWNIR region for samples containing a

large amount of water. Just as radiation in

the NIR can penetrate deeper than can

radiation in the MIDIR range, so radiation

in the SWNIR penetrates still deeper.

However, as will be shown, penetration

depth is not the only problem when

dealing with silages.

Mixed rations

Compared with forages and silages, mixed

rations have an added problem. For these

samples, in addition to all the other con-

siderations already discussed, it is neces-

sary to account for variations present in all

the ration ingredients, in developing the

calibration equations.

Theoretical Considerations

As previously discussed, because of the

way NIRS has developed, most of the

research efforts have been in the area

of practical applications and ways to

avoid or eliminate problems encountered.

Considerably less effort has been applied to

theoretical aspects of NIRS. Thus, while it

Table 9.5.NIRS resultsafor acid detergent fibre and total Kjeldahl nitrogen for dried and untreated silages
with 10 nm between wavelengths using a circular cell (Reeves and Blosser, 1991).

Calibration (n= 98) Validation (n= 47)

Grindb R^2 SEC r^2 SEP Bias

Acid detergent fibre
VIT 0.94 2.03 0.94 2.03 0.04
DRY 0.96 1.49 0.96 1.48 0.01

Crude protein
VIT 0.98 0.96 0.97 1.05 0.07
DRY 0.99 0.58 0.99 0.61 0.01

aNIRS results: R (^2) = calibration r-square; SEC = standard error of calibration; r (^2) = validation r-square; SEP =
standard error of prediction; Bias = difference in predicted and actual mean for validation set samples.
bGrinds: VIT = Vita-Mix using dry ice; DRY = ground dried material using a Wiley grinder.