Revisiting the empirical models for non-destructive grapevine shoot leaf area estimation: validation and methodological simplification using data from the cv. Encruzado

Abstract

Aiming to test the performance of the Lopes & Pinto (2005) methodology for non-destructive estimation of grapevine shoot leaf area, a sample of 100 shoots of the white cultivar Encruzado was periodically collected during 2025 season (five sampling dates) from an experimental vineyard located at Tapada da Ajuda, Lisboa. From each shoot, primary and lateral leaves were separated, imaged with a commercial camera, and their area assessed using ImageJ® software. The validation of the Lopes & Pinto model for the estimation of primary shoot leaf area (SLA1), based on an explanatory variable calculated as the mean of the largest and smallest primary leaf areas times the number of primary leaves, showed a lower mean absolute percent error (MA%E = 7.0%) and a higher modeling efficiency (EF = 0.97). Validation of the Lopes & Pinto model for lateral shoot leaf area (SLA2; n = 33) resulted in a higher MA%E (13.0%) and a lower EF (0.88) than those obtained for SLA1. To compare the goodness of fit of the Encruzado-specific models with that of the Lopes & Pinto models, the same type of regression models were computed using the Encruzado dataset. Compared with the Lopes & Pinto models, the Encruzado regression models explained a high proportion of the leaf area variability and showed very good predictive capability for both SLA1 (R² = 0.98; MA%E = 7.0; EF = 0.98) and SLA2 (R² = 0.99; MA%E = 7.9; EF = 0.97). Aiming to test the possible the possibility of reducing the number of field measured variables, the Encruzado datasets were re-analyzed using only two measured variables (number of leaves and area of the largest leaf). The resulting models, although showing a slightly lower predictive performance (SLA1: Adj. R² = 0.98; MA%E = 7.3%; SLA2: Adj. R² = 0.97; MA%E = 11.1%) than the previous ones, can still be considered highly accurate. Our results confirm the good performance of the simple and low-cost Lopes & Pinto (2005) non-destructive method for shoot leaf area estimation, and show that model accuracy increases when models are built with data from the cultivar under study. Furthermore, the procedures can be simplified by using fewer measured variables, with minimal impact on predictive performance.

References

Lopes CM and Pinto PA (2005). Easy and accurate estimation of grapevine leaf area with simple mathematical models. Vitis, 44(2), 55-61. https://doi.org/10.5073/vitis.2005.44.55-61