Determination of starch and carbohydrate in mango leaves using Vis-NIR spectroscopy

• Published in: Temas agrarios (Montería, Córdoba, Colombia) Vol. 27; no. 2; pp. 397 - 410
• Main Authors: Elisson Alves Santana, Daniel dos Santos Costa, Jose Francismar de Medeiros
• Format: Journal Article
• Language: English; Portuguese
• Published: Universidad de Córdoba 01-07-2022
• Subjects: biochemical analysis; chemometrics; mangífera indica l; non-destructive methods; potassium fertilizers
• ISSN: 2389-9182
• DOI: 10.21897/rta.v27i2.3114

Mango production presents challenges, such as the maturation of the mango branches, which, combined with good nutrition and biochemicals involved in this process, such as carbohydrate and starch favor the development of the plant. Therefore, the use of non-destructive, fast techniques to determine the levels of these components in the plant, such as spectroscopy, can optimize the analysis of these components. Therefore, this work aimed to develop predictive models for determination of starch and carbohydrate contents in “Palmer” mango leaves using vis-nir spectroscopy subjected to different potassium sources. The work was carried out in the region of San Francisco Valley, using the following steps: (1) leaf sampling; (2) spectral analysis; (3) lab determination of carbohydrate and starch contents; and (4) development of predictive regression and classification models. The predictive regression models used were Principal Components Regression (PCR) and Partial Least Squares Regression (PLSR). Supervised discriminant models were also developed to classify mango leaves according to different potassium sources used, using linear discriminant analysis (LDA). Vis-NIR spectroscopy showed low values for the non-destructive evaluation of “Palmer” mango leaves using PCR and PLSR for carbohydrate and starch prediction with R2 of 0.58 lower than the models considered excellent (R2 >0.90); The development of classification models did not allow the discrimination of different sources of potassium in “Palmer” mango leaves with an accuracy of 64.2%.