Grapevine Phenology Prediction: A Comparison of Physical and Machine Learning Models

The reduction of plant pest treatments contributes to a more sustainable agriculture. However, to be effective, the application of these treatments must be performed at the correct phenological stage of the plants. In this paper, we present the comparison of physical and ML models to predict the phenological stage of vineyards. The performance of both shows an average R2 above 0.94. However, the physical models do not generalize well and they cannot be easily improved by the inclusion of new datasets as ML models do.