Potential of visible-near infrared spectroscopy combined with chemometrics for analysis of some constituents of coffee and banana residues.

Banana (stalk, leaf, rhizome, rachis and stem) and coffee (leaf and husks) residues are promising feedstock for fuel and chemical production. In this work we show the potential of near-infrared spectroscopy (NIR) and multivariate analysis to replace reference methods in the characterization of some constituents of coffee and banana residues. The evaluated parameters were Klason lignin (KL), acid soluble lignin (ASL), total lignin (TL), extractives, moisture, ash and acid insoluble residue (AIR) contents of 104 banana residues (B) and 102 coffee (C) residues from Brazil. PLS models were built for banana (B), coffee (C) and pooled samples (B+C). The precision of NIR methodology was better (p<0.05) than the reference method for almost all the parameters, being worse for moisture. With the exception of ash (B and C) and ASL (C) content, which was predicted poorly (R(2)<0.80), the models for all the analytes exhibited R(2)>0.80. The range error ratios varied from 4.5 to 16.0. Based on the results of external validation, the statistical tests and figures of merit, NIR spectroscopy proved to be useful for chemical prediction of banana and coffee residues and can be used as a faster and more economical alternative to the standard methodologies.