Regression model for leaf area estimation in Ficus carica L.

Leaf area (LA) is an important parameter to elucidate the relationship between plant growth and environment. LA can be measured by direct or indirect methods. In the current experiment, a LA estimation model was developed for common fig (Ficus carica L.) using linear measurements of leaf length (L) and leaf width (W). Leaves from six F. carica cultivars: 'Calabrese Nero', Cipressotto', 'Columbro Nero', Natalina', 'Papa', and San Mango' were used to calibrate the model. Regression analysis of LA vs. L and W revealed several models that could be used for estimating the area of individual F. carica leaves. A linear model having L and W as independent variables provided the most accurate estimate: highest R2 (>0.95), smallest mean square errors (MSE), smallest prediction sum of squares (PRESS), and a PRESS value reasonably close to the error sum of squares (SSE). However, this model required double time for leaf measurement. The model having W as independent variable exhibited also a high accuracy and precision in estimating individual F. carica LA. We preferred this model because of its simplicity and convenience, as it only involves one variable. Model validation was carried out using an independent data set coming from another F. carica cultivar ('Dottato Bianco'). Correlation coefficients showed a strong relationship between predicted and observed leaf area with an underestimation of 6.6%. This model can be reliably adopted without recalibration to estimate LA of F. carica cultivars, non-destructively.