*3.3. Characteristic Wavelengths*

S-G(7, 5)&

Fewer than 100 wavelengths were required to simplify the SCC evaluating models. The changing processes of RMSEV with different wavelength candidate subsets, which were chosen by SPA or MCUVE, are shown in Figure 3a,b, respectively. Using the combination of (SPA, PLSR) or (MCUVE, PLSR), characteristic wavelengths were selected (57 and 83, respectively). The minimum RMSEVs of SPA and MCUVE were 0.0692% and 0.0685%, respectively.

**Figure 3.** Changing processes of RMSEV with different wavelengths. (**a**) SPA; (**b**) MCUVE.

As shown in Figure 4a, 46 characteristic wavelengths distributed densely between 846.8 and 940.6 nm were obtained for the first selection method, while 10 characteristic wavelengths were scattered between 498 and 750 nm and 1 characteristic wavelength was located at 997.1 nm. Characteristic wavelengths selected by the second method were distributed mainly in the ranges of 757.7–796.7 nm, 828.3–847.5 nm, 866.8–910.3 nm, and 952.9–1006.9 nm, as shown in Figure 4b. The two methods both selected characteristic wavelengths in the range of 828.3–910.3 nm, which were correlated with the third overtone stretch of O-H and C-H functional groups.

**Figure 4.** Distribution of characteristic wavelengths. (**a**) SPA; (**b**) MCUVE.
