Rapid assessment of Italian polyfloral honey overall quality index based on non-destructive spectral sensors.

The quality of polyfloral honey is based on different physicochemical and sensorial characteristics such as sugars, pH, colour, texture, odour and flavour, appealing to varied consumer preferences. Traditional honey quality assessment methods, such as solvent extraction and laboratory analyses, are time-consuming, costly, and require chemical reagents. This study investigated the efficacy of a Near Infra-Red (NIR)-based device for rapid honey quality evaluation. A total of 215 Italian polyfloral honey samples collected from different beekeepers in 2022 were analyzed for parameters like moisture, water activity (aw), conductivity, pH, free acidity, diastase index, sugars (maltose, glucose, fructose), and rheological traits (colour, odour, taste and toughness). Per each trait, samples were assigned to quality classes according to a tertiles classification (1–3; low to high) based on standard thresholds [1]. An overall quality index (QI) was determined as the sum of the quality class scores for each individual trait. NIR spectra (850–2500 nm) were acquired in transmittance mode using a FOSS DS2500 spectroscopic instrument. The device operates within a scanning range of 850 to 2500 nm, with 0.5 nm intervals, and is equipped with a small rotating ring cup featuring a gold reflector with a depth of 0.5 mm. Using WinISI 4 V4.10.0.15326 (FOSS Analytical A/S, Hillerød, Denmark), a partial least squares discriminant analysis (PLS-DA) was applied to the collected spectral data. Finally, a confusion matrix was utilized to derive metrics, including sensitivity, specificity, accuracy, and the Matthews correlation coefficient (MCC), which collectively evaluated the PLS-DA model’s effectiveness in discriminating honey quality. Based on the recorded reference analysis, the overall QI values were categorized into three tertiles: below 27 as low (LQI), between 27 and 31 as medium (MQI), and above 31 as high (HQI). Results showed sensitivity, specificity, accuracy, and MCC values of 0.71–0.71, 0.82–0.88, 0.79–0.82, and 0.50–0.61, respectively. MQI class was correctly assigned in 71% of samples, with minimal misclassification into nearest groups (16% to LQI, 13% to HQI). Similarly, LQI and HQI groups were correctly classified in 72% and 71% of cases, respectively, with most misclassifications occurring also in the nearest merit class. The findings demonstrate that NIR spectroscopy effectively discriminates among honey quality classes for Italian polyfloral honey. It provides a rapid, non-destructive method for segregating high-quality polyfloral honey in productive Italian ecological areas.