Bark quantification and quality classification in Pinus spp. pellets: a non-destructive approach using nir spectroscopy and cielab colorimetry

Abstract

Blending bark and sawdust from Pinus spp. to produce pellets adds value to forest residues, but it systematically alters the product's density, mechanical strength, and energy content. Conventional quality control methods are destructive and slow, unable to verify bark proportion in real time or to anticipate deviations before commercialization. Near-infrared (NIR) spectroscopy and CIELab colorimetry are non-destructive alternatives, but calibrations for bark–sawdust mixtures across the full gradient are lacking. This dissertation evaluated the potential of both techniques to quantify bark content and to classify the physical, mechanical, and energy quality of pellets produced under eleven treatments (T0–T10, 10% bark increments, 0–100%). Physical-mechanical and energy properties varied among treatments (p < 0.001). The best balance between durability and energy density occurred between 40% and 60% sawdust, but no treatment reached the minimum mechanical durability threshold set by ISO 17225-6 (class B, ≥96%). Through NIR spectroscopy, it was possible to quantify bark content with high accuracy in both the loose biomass (non-densified mixture) and the pellets (R²p between 0.997 and 0.998; RPDp between 18.92 and 21.84); PLS-DA models classified bark content into high, medium, and low categories with accuracies between 97.98% and 99.60% (highest for pellets with first-derivative preprocessing). Physical-mechanical and energy properties were predicted more reliably from the pellet spectrum (R²p 0.710–0.974) than from the loose biomass (R²p 0.513–0.807); moisture content was the property predicted with the highest accuracy, while higher heating value and energy density had the lowest. In colorimetry, the b* coordinate showed the strongest association with energy density (r = −0.89; R² = 0.79). Three classes, defined by bark proportion, showed distinct profiles and their own colorimetric ranges: L* showed no overlap between classes, and a* was the coordinate most closely associated with bark content. Class II (40–70% bark) achieved the highest energy density (13.47–14.32 GJ m⁻³). A principal component analysis integrating these variables confirmed the separation into three groups and explained 86.1% of the variance across the first three components. NIR spectra, combined with PLS-DA and PLS-R models, provided the highest accuracy for quantifying bark proportion in mixtures and pellets of Pinus spp.; CIELab colorimetry, using a portable colorimeter, allowed rapid, non-destructive classification of pellet energy quality. The two techniques are complementary and can be incorporated into the industrial quality control of bark-and-sawdust pellets.