Stephania Imbachi-Ordonez and Gillian Eggleston
Audubon Sugar Institute, LSU AgCenter, St. Gabriel, Louisiana, USA; simbachiordonez@agcenter.lsu.edu
In many sugarcane-producing regions, payment systems often fail to account for extraneous matter (EM), such as leaves and soil, which are harvested along with the sugarcane stalks. This oversight fails to incentivize growers to deliver higher-quality cane and negatively impacts factory efficiency. The lack of practical methodologies to quantify EM further complicates this issue. Near-Infrared Spectroscopy (NIR), a rapid and non-destructive analytical technique, offers a potential solution. The application of NIR for quantifying EM in sugarcane, specifically leaves and soil, was evaluated to support its future inclusion in the Louisiana (LA) cane-payment system. Mixtures of clean cane, soil, and leaves (green and brown) of known concentrations were prepared using the most representative sugarcane varieties and soil types in LA. The cane-soil-leaves mixtures were then used to develop NIR calibration models for predicting leaf content, while core shredded cane samples were used to calibrate soil content based on incinerated ash analysis. Partial least squares regression (PLSR) models with k-fold cross-validation were applied to relate NIR spectra to reference values. The NIR calibration for soil content, based on incinerated ash analysis, achieved an R² of 92.86% with a Root Mean Square Error of Cross-Validation (RMSECV) of 0.75. NIR successfully predicted leaves (R² = 88.55%, RMSECV = 2.42). We demonstrate the practical application of NIR spectroscopy for the rapid and accurate quantification of EM in shredded sugarcane, particularly for soil and total leaves. Machine-learning techniques will be applied to refine these models and improve prediction accuracy.