Visible and near infrared spectroscopy for the analysis of nutrients in sugarcane plant tissue for panela production
Near Infrared Reflectance Spectroscopy (NIRS) is a fast, multiparametric, environmentally friendly, low-cost, and highly accurate technology for the analysis of components in food, soil, and agriculture. The purpose of this study was to generate NIRS calibration models for the prediction of nutrients in plant tissue of sugarcane to panela production cultivated in the Hoya del Río Suárez region. A total of 416 tissue samples were scanned in Vis-NIR spectral segment. Chemometric analysis was performed with the WinISI V4.10 software applying modified partial least squares regression with cross-validation. Four models with different mathematical treatments were evaluated, and the performance of calibrations was made through external validation analyzing the goodness-of-fit measures as prediction determination coefficient, standard error of the bias-adjusted prediction, and residual predictive deviation. The results showed that the calibration model for N had the highest predictive power. For macronutrients, the calibrations with the best predictive power were for P and K, and micronutrients for B, while Cu presented the lowest predictive power. Adequate models were found for the prediction of N, Ca, and P. In the case of the other nutrients, it is recommended to expand the calibration set.
| Main Authors: | , , , , , |
|---|---|
| Format: | Digital revista |
| Language: | spa |
| Published: |
Universidad de Ciencias Aplicadas y Ambientales U.D.C.A
2023
|
| Online Access: | https://revistas.udca.edu.co/index.php/ruadc/article/view/2062 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Near Infrared Reflectance Spectroscopy (NIRS) is a fast, multiparametric, environmentally friendly, low-cost, and highly accurate technology for the analysis of components in food, soil, and agriculture. The purpose of this study was to generate NIRS calibration models for the prediction of nutrients in plant tissue of sugarcane to panela production cultivated in the Hoya del Río Suárez region. A total of 416 tissue samples were scanned in Vis-NIR spectral segment. Chemometric analysis was performed with the WinISI V4.10 software applying modified partial least squares regression with cross-validation. Four models with different mathematical treatments were evaluated, and the performance of calibrations was made through external validation analyzing the goodness-of-fit measures as prediction determination coefficient, standard error of the bias-adjusted prediction, and residual predictive deviation. The results showed that the calibration model for N had the highest predictive power. For macronutrients, the calibrations with the best predictive power were for P and K, and micronutrients for B, while Cu presented the lowest predictive power. Adequate models were found for the prediction of N, Ca, and P. In the case of the other nutrients, it is recommended to expand the calibration set. |
|---|