The 2012 ASHS Annual Conference

Açaí (Euterpe oleracea Mart.) fruit has a dark-purple skin colour due to the presence of anthocyanins which have excellent functional properties, high antioxidant capacity. The standard methods to quantify anthocyanin are destructive, time consuming, generate chemical residues and sometimes require specialized procedures. Therefore non-invasive and/or nondestructive techniques have been used to determine quality parameter of fruits and vegetables e.g. the near infrared spectroscopy (NIRS). Although the apparent simplicity of NIRS the use of fruit samples from different batches is probably the most important factor that may affect model performance, as the fruit matrix may be subjected to within-tree, within-orchard, fruit age and seasonal variability. This study describes the model performance of total anthocyanin content prediction of intact açaí fruit (E. oleracea Mart.) using NIR diffuse reflectance spectroscopy based on different validation procedures. The models we have developed were obtained from açaí fruits samples collected during 2010–11 at 3 harvest periods and at 4 growing regions. The spectra were pretreated using standard normal variate (SNV), de-trend transformation and first derivative (Savitzky-Golay). Calibrations were developed using partial least squares (PLS) regression. Model performance was evaluated based on the values of root mean square error for prediction (RMSEP) and coefficient of determination (R²) obtained from different validation fruit samples, as such: i) random cross validation method; ii) one third of used spectra dataset; iii) external validation of an independent dataset; and iv) external validation of an independent dataset of fruits coming from Pará State. The model constructed using random cross validation method lead to a RMSEP of 0.46% with 5 latent variables (LVs). More robust model was obtained when one third of the spectra dataset was used (RMSEP 0.15%, R² 0.96), but the LVs increased to 15. Independent dataset resulted in a less robust model (RMSEP 0.23%, R² 0.90, 12 LV) compared to one third of the dataset. Course model performance was obtained when fruits from Pará State independent dataset was used as test matrix (RMSEP 0.77%) as the R² was not adjusted (0.04). The NIRS can be successfully used to predict anthocyanin in açaí fruits as a nondestructive method, however more dataset of fruits from different batches is necessary to reduce RMSEP and improve calibration model prediction accuracy and robustness.

Abstracts:

• Graf. açaí external validation .jpg (40.8KB) - Abstract Text