364
T. Ma et al.
has been shown to play a large role in the protection of living tree [3]. The first
NIR works were mostly focused on evaluating wood chemistry directly, especially
the cellulose content [4–6] and then, shift to estimate wood lignin and extractives.
Da Silva et al. presented an assessment of total phenolic compounds and extractive
contents of mahogany wood rapidly by NIRS [7]. He and Hu indicated the benefits of FT-NIR to predict the lignin and extractive content of different wood species
[8]. The validation results confirm that the selection of relevant wavenumbers and
suitable data preprocessing methods produced values within tolerance levels. Lepoittevin et al. indicated that it is important to remove extractives before NIR spectra
collection for the prediction of other wood chemistry traits [9]. Uner et al. utilized
the genetic inverse least squares method for constructing the calibration models of
lignin and extractive contents in milled Turkish pine wood samples. The standard
error of calibration (SEC) and standard error of prediction (SEP) were 0.35% and
2.40%, respectively [10].
16.2.2 Wood Moisture Content
The molar absorption of water at the NIR range is 1/1000–1/10,000 compared to that
of IR region [11]. Nevertheless, as the NIR range has rich light absorbance information of oxygen and hydrogen (O–H) structures, many researchers have invested
NIRS to predict water within wood. In general, the state of water within wood can
be categorized as either free or bound water. Here, free water is defined as liquid
water located in the lumens and intercellular spaces of wood but without a chemical
bond with the wood cell wall; whereas, the water attached by intermolecular forces
between the major chemical components of wood cell walls is considered as bound
water, which has profound effects on wood physical properties. For MC by mass
(i.e., both free and bound water), Watanabe et al. compared the accuracy of NIRS
with a commercial capacitance-type moisture meter for greenwood sorting purposes.
Their experimental results showed the performance of NIR approach was better than
the capacitance-type at predicting high moisture wood samples. Besides, compared
to the capacitance-type moisture meter, the NIR method also has the advantage of
measuring MC without the need for density correction [12]. However, Defo et al.
suggested that NIR spectrometer may be less useful for the lumbers with significant gradients between core and surface layers [13]. For this limitation, Tham et al.
recently highlighted the potential of NIRS combined with an industrial MC capacitance meter to predict MC from greenwood to oven-dried conditions [14]. Experimental results showed a good prediction accuracy (coefficient of determination (R
2 )
= 0.80, root mean square error of cross-validation (RMSECV) = 25.70%, and the
ratio of percentage deviation (RPD) = 2.22) could be achieved at various sample
thicknesses and wood species without density correction. It suggests that NIRS can
be assisted by other techniques with higher transmission abilities, when measuring
thick wood samples such as timber and lumber wood.
T. Ma et al.
has been shown to play a large role in the protection of living tree [3]. The first
NIR works were mostly focused on evaluating wood chemistry directly, especially
the cellulose content [4–6] and then, shift to estimate wood lignin and extractives.
Da Silva et al. presented an assessment of total phenolic compounds and extractive
contents of mahogany wood rapidly by NIRS [7]. He and Hu indicated the benefits of FT-NIR to predict the lignin and extractive content of different wood species
[8]. The validation results confirm that the selection of relevant wavenumbers and
suitable data preprocessing methods produced values within tolerance levels. Lepoittevin et al. indicated that it is important to remove extractives before NIR spectra
collection for the prediction of other wood chemistry traits [9]. Uner et al. utilized
the genetic inverse least squares method for constructing the calibration models of
lignin and extractive contents in milled Turkish pine wood samples. The standard
error of calibration (SEC) and standard error of prediction (SEP) were 0.35% and
2.40%, respectively [10].
16.2.2 Wood Moisture Content
The molar absorption of water at the NIR range is 1/1000–1/10,000 compared to that
of IR region [11]. Nevertheless, as the NIR range has rich light absorbance information of oxygen and hydrogen (O–H) structures, many researchers have invested
NIRS to predict water within wood. In general, the state of water within wood can
be categorized as either free or bound water. Here, free water is defined as liquid
water located in the lumens and intercellular spaces of wood but without a chemical
bond with the wood cell wall; whereas, the water attached by intermolecular forces
between the major chemical components of wood cell walls is considered as bound
water, which has profound effects on wood physical properties. For MC by mass
(i.e., both free and bound water), Watanabe et al. compared the accuracy of NIRS
with a commercial capacitance-type moisture meter for greenwood sorting purposes.
Their experimental results showed the performance of NIR approach was better than
the capacitance-type at predicting high moisture wood samples. Besides, compared
to the capacitance-type moisture meter, the NIR method also has the advantage of
measuring MC without the need for density correction [12]. However, Defo et al.
suggested that NIR spectrometer may be less useful for the lumbers with significant gradients between core and surface layers [13]. For this limitation, Tham et al.
recently highlighted the potential of NIRS combined with an industrial MC capacitance meter to predict MC from greenwood to oven-dried conditions [14]. Experimental results showed a good prediction accuracy (coefficient of determination (R
2 )
= 0.80, root mean square error of cross-validation (RMSECV) = 25.70%, and the
ratio of percentage deviation (RPD) = 2.22) could be achieved at various sample
thicknesses and wood species without density correction. It suggests that NIRS can
be assisted by other techniques with higher transmission abilities, when measuring
thick wood samples such as timber and lumber wood.