LDA and PLS-DA were applied to sharpen the separation between
samples and build classification models. Seventy-nine samples
(27 from ‘‘Guyuelongshan’’, 27 from ‘‘Kuaijishan’’ and 25 from
‘‘Pagoda’’) were selected as the calibration sample set, and the
remaining 38 samples (13 from ‘‘Guyuelongshan’’, 13 from ‘‘Kuaijishan’’
and 12 from ‘‘Pagoda’’) were used as the prediction sample
set.