Uncertainties in predicting the medium-term future occurrence of grapevine budburst in Europe: the role of phenological models and climate scenarios

Budbreak plays an important role in the grapevine growing cycle and temperature is its main driver. Therefore, phenological models use two temperature-based approaches to simulate budbreak: a chilling-forcing scheme, which describes either endo- and eco-dormancy periods, or a forcing-approach, which exclusively simulates the eco-dormancy period. Both approaches are able to estimate budbreak under current temperature conditions, but they diverge under future climate forecasts. Additional divergences in phenological estimation are driven by climate as simulated by different global and regional circulation models and GHG concentration scenarios. Thus, this study explored the sources of uncertainty in budbreak estimation across Europe in a historical baseline (1976-2005) and near-future (2026-2055) climate. The experimental design comprised six phenological models calibrated for eight distinct grapevine varieties. These phenological models were applied to one historical and two future representative concentration pathways using different combinations of regional and global climate models according to data availability. In total, 25 experiments were performed for the historical period and 35 for the near-term future (18 and 17 for RCP2.6 and RCP4.5 scenarios respectively). The results showed different spatial domains of uncertainty across Europe. The total uncertainty in estimating budbreak was low in Central Europe and increased outside these regions in both the historical and future periods. Specifically, the uncertainty in the baseline period was mainly related to the phenological models (~94.7 %) with slight spatial differences across the study area. In the future period, Central Europe was characterised by high uncertainties due to the climate models (~40 %). Outside these regions, uncertainty increased due to the phenological models, the highest uncertainties being associated with the Mediterranean basin for the cold-adapted varieties, while the north/northeast regions showed the highest uncertainties for the warm-adapted varieties. High temperatures resulted in low daily chilling rates for BRIN, while only certain temperatures positively contributed for UNIFIED and UNICHILL. Additionally, low temperatures did not accumulate forcing units for any phenological model, while an increase in temperature led to a linear (GDD, Richardson-BRIN until a threshold) or parabolic (WANG, UNIFORC) increase in the daily forcing unit rate. These differences limit the use of the phenological models, which will need to be taken into account when applying these models in their application in different environments in the future.