Total soluble solids (TSS) is a key variable taken into account in determining optimal grape maturity for harvest. In this work, partial least square (PLS) regression models were developed to estimate TSS content for Godello, Verdejo (white), Mencía, and Tempranillo (red) grape varieties based on diffuse spectroscopy measurements. To identify the most suitable spectral range for TSS prediction, the regression models were calibrated for four datasets that included the following spectral ranges: 400–700 nm (visible), 701–1000 nm (near infrared), 1001–2500 nm (short wave infrared) and 400–2500 nm (the entire spectral range). We also tested the standard normal variate transformation technique. Leave-one-out cross-validation was implemented to evaluate the regression models, using the root mean square error (RMSE), coefficient of determination (R²), ratio of performance to deviation (RPD), and the number of factors (F) as evaluation metrics. The regression models for the red varieties were generally more accurate than the models of those for the white varieties. The best regression model was obtained for Mencía (red): R² = 0.72, RMSE = 0.55 °Brix, RPD = 1.87, and factors n = 7. For white grapes, the best result was achieved for Godello: R² = 0.75, RMSE = 0.98 °Brix, RPD = 1.97, and factors n = 7. The methodology used and the results obtained show that it is possible to estimate TSS content in grapes using diffuse spectroscopy and regression models that use reflectance values as predictor variables. Spectroscopy is a non-invasive and efficient technique for determining optimal grape maturity for harvest.

While populations of the Asian chestnut gall wasp (Dryocosmus kuriphilus Yasumatsu), an invasive pest affecting the European chestnut (Castanea sativa Miller), have started to be controlled biologically, this pest still conditions chestnut tree development. With the aim of assessing plant health status as a means of monitoring gall wasp infestation, we used a field spectroradiometer to collect data from leaves taken from 83 trees in two chestnut orchards. We calculated characteristic spectral signatures for pest infestation, and after training and validation, developed classifiers to distinguish between different infestation levels. Several partial least square discriminant analysis (PLS-DA) and random forest (RF) models were fitted with reflectance and transformed values to obtain characteristic curves reflecting infestation. Four wavelengths (560 nm, 680 nm, 1400 nm, and 1935 nm) were identified as showing the greatest differences between curves. The best overall accuracy (69.23%) was achieved by an RF model fitted with reflectance transformed values. Lower overall accuracy (26.92%) was achieved in distinguishing between infestation levels. In conclusion, while more specific differences in infestation levels were not detectable, our method successfully discriminated between gall absence and presence.