A simple and accurate model for the non-destructive estimation of leaf areas in genotypes of Plumeria rubra L.

Abstract

SummaryLeaf area (LA) is an important parameter in many plant modelling studies.There is a need for a simple, accurate and non-destructive model to predict LA in physiological experiments in which destructive LA sampling is not allowed (e.g., rare plants or genetically segregating populations). In this study, a model for LA estimation was developed for Plumeria rubra L. using simple linear measurements of leaf length (L) and maximum width (W). Two experiments were carried out, one in 2012 (on seven commercial genotypes: ‘California Sunset’, ‘Christina’, ‘Elsie’, ‘Gina’, ‘J.J. Mini White’, ‘Panittas Red’, and ‘Pixie Dust’) and one in 2013 (on one genotype ‘Divine’) under greenhouse conditions, to test whether a model could be developed to estimate LAs of P. rubra across genotypes. Regression analysis of LA vs. L and/or W revealed several models that could be used to estimate the area of individual Plumeria leaves.To develop a model to estimate individual LA values accurately for all genotypes of P. rubra, measurements of both L and W should be included. The proposed linear model [LA = 4.15 _ 0.66 (L _ W)] was adopted for its accuracy, highest R2 value (> 0.96), smallest mean square error (MSE), and smallest predicted residual error in the sum of squares (PRESS), and whose PRESS value was close to the sum of squares for error (SSE).Validation of the model using L _W measurements of leaves was carried out on an independent dataset derived from another genotype (‘Divine’) in 2013. Correlation coefficients showed that there were strong relationships between predicted LAs and observed LAs, with an over-estimation of 4.0% in the prediction. This model can therefore be adopted reliably to estimate LAs in P. rubra, non-destructively.