Development of predictive models for simulating vegetative and reproductive phenology in European hazelnut trees...

In Chile, European hazelnut production has expanded substantially in Mediterranean and temperate regions in recent years. However, these areas are experiencing increasing temperature variability and more frequent heat waves due to climate change. Such climatic changes modify tree phenology, altering the onset and progression of endodormancy and ecodormancy, which ultimately affects nut quality and yield. In this context, phenological models are essential for developing decision-support tools in hazelnut orchard management, including pest and disease control, irrigation, and fertilization. The study aimed to develop logistic and monomolecular models based on growing degree days (GDD) to simulate vegetative (VPS) and reproductive (RPS) phenological stages of European hazelnut (Tonda di Giffoni' and 'Barcelona') under Mediterranean and temperate agroclimatic conditions. Additionally, a stochastic framework was implemented to evaluate uncertainty in phenological predictions. Phenological observations, based on the Giardini Italian phenological stages (GFI), together with agroclimatological data, were collected from three commercial orchards in the Maule, Ñuble, and Araucanía Regions of Chile during four consecutive growing seasons (2019-2020, 2020-2021, 2021-2022, and 2022-2023). For the development of the GDD-based models, the GFI code was transformed into a numerical phenological scale by assigning consecutive integers for mathematical representation. For both cultivars, the logistic and monomolecular models based on GDD for simulating the phenological scale (PS) achieved coefficients of determination (R²) between 0.80 and 0.94. Validation further demonstrated that the phenological models predicted PS with root mean square error (RMSE) values ranging from 0.43 to 1.08 and model efficiency (EF) values between 0.94 and 0.99. Finally, the stochastic approach indicated the range of potential outcomes that can occur due to random or unpredictable errors.